Age–incidence patterns of primary CNS tumors in children, adolescents, and adults in England

Abstract

Around 25% of all tumors in those 0–14 years of age and 9% in those 15–24 years of age involve the CNS. They are the most common cause of cancer-related deaths in both age groups. In adults 25–84 years of age, the proportion of CNS tumors is 2%; 5-year overall survival is 10%–15%; and survivors have considerable morbidity. Comprehensive up-to-date population-based incidence data on these tumors are lacking. We present incidence rates for primary CNS tumors based on data derived from the high-quality national cancer registration system in England. A total of 54,336 CNS tumors of malignant, benign, and uncertain behavior were registered across the whole of England from 1995 through 2003. The age-standardized rates for all ages (0–84 years) was 9.21 per 100,000 person-years. This is higher than previously reported for England because it includes nonmalignant CNS tumors and hence gives a more accurate picture of burden of disease. The age-standardized rates for those 0–14 years of age, 15–24 years of age, and 25–84 years of age were 3.56, 3.26, and 14.57 per 100,000 person-years, respectively. In this article, we describe the changing patterns in the epidemiology of primary CNS tumors in these three age groups with respect to sex, tumor behavior, and histology using the current WHO classification. This information will provide a reference for future studies nationally and internationally and make comparisons relevant and meaningful.

Previous data from regional registries show that, overall, 1.5%–1.9% of all cancers registered in England are tumors of the CNS, including brain and spinal cord tumors.¹ Although CNS tumors are predominantly a disease of old age, the proportion of CNS tumors among all cancers falls significantly with increasing age. CNS tumors represent 24.5% of all tumors in those 0–14 years of age² and 8.9% in those 15–24 years of age.³ Their importance as a health problem in young people is further highlighted by the fact that CNS tumors are the most common cause of cancer-related deaths in both the 0- to 14-year age group² and the 15- to 24-year age group.⁴ Even histologically benign tumors can be life-threatening because of their space-occupying effects within the cranium, local infiltration, and for some, a tendency to undergo malignant transformation over time.⁵ There is also significant morbidity both from the disease and from the treatment required, with varying degrees of physical, cognitive, neurological, endocrinological, and other deficits in survivors resulting in significant handicap and diminished quality of life.⁶

Establishing accurate incidence rates for these tumors is a challenge not only because they are a very heterogeneous group with more than 100 distinct pathological entities, but also because of variations in registration practice, changes in classification, and improvements in neurodiagnostic techniques over time. Previously, in Britain, studies by morphological type have been based on regional registry data only or cases ascertained from hospitals.⁷ There have been no national studies describing the epidemiology of CNS tumors in detail in adolescents and adults. Studies from Norway⁸ and Japan⁹ analyzed national data sets across all ages, but the diagnostic classifications used were historical and differed (1979 and 1993 WHO CNS tumor classification^10,11 in the study from Norway and the 1945 International Union against Cancer classification¹² in the study from Japan), which does not permit easy comparison. Similarly, the Automated Childhood Cancer Information System (ACCIS) has reported Europe-wide CNS tumor incidence data in children and has used the International Classification of Childhood Cancer (ICCC),¹³ which is morphology based but is not suited to older ages. The Central Brain Tumor Registry of the United States (CBTRUS) has published incidence rates from 1998–2002¹⁴ based on the 2000 WHO CNS tumor classification,¹⁵ but these are from 18 state cancer registries and cover only 32% of the U.S. population. There is also huge variability in the reporting of tumors among U.S. states, with the percentage of nonmalignant tumors varying from 27% to 60% of overall CNS tumors. Also, because data are collected from each registry without a unique identifier, there is the possibility of duplicate registration.

Cancer registration is conducted by eight regional cancer registries in England, and the essential features of the system of registration have remained unchanged for more than 30 years.¹⁶ England has a high degree of case ascertainment, and reviews have shown that registry records are largely complete, accurate, and reliable.¹⁶ Notification of cancer registrations to the National Health Service Information Centre allows completeness of registration and eliminates duplication. Data on CNS tumors obtained from these registries have been grouped using the current WHO¹⁵ classification. We present here incidence rates of CNS tumors for ages 0–14 years, 15–24 years, and 25–84 years for the whole of England during the period 1995–2003. We describe the differences in the site and pathology distributions of CNS tumors in these age ranges. This will allow us to better understand the changes with age in this large, heterogeneous collection of tumors. Use of the current WHO classification will make comparisons across geographical areas as well as over time more meaningful in the future.

Materials and Methods

Source of Data

Cancer registration in England is carried out by a network of eight population-based regional registries. Registration is coordinated by the Office for National Statistics in London, which maintains the national cancer registry covering all age groups. Anonymized individual-level national cancer registration data were obtained from the Office for National Statistics for all CNS tumors (tumor at any of the following sites: brain, meninges, spinal cord, cranial nerves, other parts of the CNS, pituitary, and pineal glands) of malignant, benign, and uncertain behavior newly diagnosed between 1995 and 2003. Information was available on the year of diagnosis, age at diagnosis, sex of patient, primary site code, morphology code, and behavior code. Individual-level data on ethnicity was not available. National population estimates by single year of age, gender, and calendar year were supplied by the Population Estimates Unit, Office for National Statistics.

Classification

The data obtained were classified into diagnostic groups to match the WHO 2000 classification on the basis of International Classification of Diseases for Oncology (ICD-O) M and T codes.¹⁵ Modifications had to be made to make the classification more comprehensive by including pituitary tumors, not otherwise specified (NOS), and unspecified CNS tumors. This is consistent with the modified version of the new WHO 2000 classification used by CBTRUS.¹⁴ We excluded metastatic tumors and tumors that were of uncertain primary/metastatic status. Also excluded were CNS lymphomas, hemopoietic neoplasms, mesenchymal nonmeningothelial tumors, and olfactory tumors. The final version of our classification is given in appendix A, where any departure from the current WHO classification are in bold and italicized.

Statistical Methods

Age- and sex-specific incidence rates were calculated and expressed per 100,000 person-years. Histology and site-specific incidence rates were also calculated for three different age groups: 0–14 years, 15–24 years, and 25–84 years. Those older than 85 years of age were excluded because of possible underascertainment and lower specificity in diagnosis. All rates were adjusted to the world standard population¹ using direct methods, except where specifically stated. SPSS version 15.0 (SPSS Science, Inc., Chicago, IL, USA) and Excel version 2003 (Microsoft Corp., Redmond, WA, USA) were used to analyze the data and produce tables and graphs.

Results

Overall Incidence

During the period 1995–2003, 54,336 primary CNS tumors of malignant, benign, and uncertain behavior located in the brain, meninges, spinal cord, cranial nerves, other parts of the CNS, and pituitary and pineal glands were registered in England in persons 0–84 years of age, which gives an annual average of just more than 6,000 new cases. The population covered was all individuals between 0 and 84 years of age in England from 1995 through 2003, which equates to 432 million person-years. There were 28,069 male cases (51.7%) and 26,267 female. The overall incidence rate was 9.21 per 100,000 person-years, and the male and female incidence rates were 9.96 and 8.52 per 100,000 person-years, respectively, giving a male-to-female ratio of 1.17:1.

Age-Specific Incidence

The age-specific incidence rates for primary CNS tumors are shown in Fig. 1. Peak incidence was seen in the 75- to 79-year age group for males and females. The number of 0- to 14-year-olds with primary CNS tumors was 2,959, with an annual average of approximately 330 new cases and an incidence rate of 3.56 per 100,000 person-years. The male and female incidence rates for the 0- to 14-year age group were 3.72 and 3.39 per 100,000 person-years, respectively. There were 1,764 cases among persons 15–24 years of age, with an annual average just below 200 new cases and an incidence rate of 3.26 per 100,000 person-years. The male and female incidence rates for the 15- to 24-year age group were 3.47 and 3.04 per 100,000 person-years, respectively. There were 49,612 cases, with an annual average of about 5,500 new cases, in the 25- to 84-year age group; the incidence rate was 14.57 per 100,000 person-years, and the male and female incidence rates for the same age group were 15.86 and 13.40 per 100,000 person-years, respectively.

Distribution by Tumor Behavior

Most primary CNS tumors in the 0- to 84-year age group were malignant (60%), and the overall incidence rate for malignant tumors for all ages was 5.64 per 100,000 person-years. The incidence rates for tumors of benign and uncertain behavior were 2.78 and 0.79 per 100,000 person-years, respectively. Tumors of malignant behavior decreased in proportion with increasing age, while tumors of benign behavior increased in proportion (Fig. 2). Within the malignant group, the astrocytomas in those 0–14 years of age were mainly low grade (WHO grade I and II), and those in the 25- to 84-year age group were high grade (WHO grade III and IV), while in those 15–24 years of age there was an equal proportion of low- and high-grade astrocytomas.

Distribution by Site

The distribution of tumors by primary site within the CNS for age groups 0–14 years, 15–24 years, and 25–84 years is shown in Fig. 3. Tumors located in infratentorial brain decreased in proportion with increasing age, while tumors located in supratentorial brain and meninges increased in proportion. Tumors of the pituitary and pineal glands and of the craniopharyngeal duct were relatively higher in proportion in the 15- to 24-year age group than at other ages.

Distribution by Histology

The distribution of tumors by main histology groups within the CNS for age groups 0–14 years, 15–24 years, and 25–84 years is shown in Fig. 4. Tumors of neuroepithelial tissue decreased in proportion with increasing age, while meningeal and unspecified tumors increased in proportion.

The data on median age at diagnosis, incidence rates, and male-to-female ratio are shown in Tables 1 and 2. There was an overall male preponderance, but meningiomas showed a strong female preponderance (p < 0.0001), and they were twice as common in adult females compared with adult males. The most common specified tumors registered in the 0- to 14-year age group were pilocytic astrocytomas, medulloblastomas, and ependymal tumors. Craniopharyngioma was the most common nonneuroepithelial primary CNS tumor in children. Tumors with their peak incidence rates in those younger than 1 year of age were choroid plexus tumors, gangliogliomas, supratentorial primitive neuroectodermal tumors, and teratomas (data not shown). Tumors with their peak incidence rates in those 1–14 years of age were pilocytic astrocytomas, subependymal giant cell astrocytomas, anaplastic ependymomas (some of them may actually be ependymoblastomas, because they have the same morphological code, 9,392/3¹⁷), medulloblastomas, and craniopharyngiomas. The most common specified tumors registered in the 15- to 24-year age group were pituitary tumors, pilocytic astrocytomas, and nerve sheath tumors, while tumors with their peak incidence rates in that age group were pleomorphic xanthoastrocytomas, neurocytomas, and germinomas. Overall, the most common specified tumors registered in the 25- to 84-year age group were glioblastoma multiforme, meningiomas (94.5% nonmalignant), and pituitary tumors. Neuroepithelial tumors peaking in this age group were specified diffuse astrocytomas (peak incidence rate at 50–54 years of age), anaplastic astrocytomas, glioblastoma multiforme (peak incidence rate at 65–69 years of age), oligodendrogliomas (peak incidence rate at 50–54 years of age), anaplastic oligodendrogliomas (peak incidence rate at 55–59 years of age), mixed gliomas (peak incidence rate at 50–54 years of age), myxopapillary ependymomas, and subependymomas. Nonneuroepithelial tumors peaking in this age group included nerve sheath tumors (peak incidence rate at 60–64 years of age), meningiomas (peak incidence rate at 80–84 years of age), pituitary tumors (peak incidence rate at 65–69 years of age), and hemangioblastomas.

Discussion

We believe that this large, comprehensive, and up-to-date analysis of incidence data accurately reflects the incidence of primary CNS tumors in England. A total of 54,336 CNS tumors of benign, uncertain, and definite malignant behavior were registered across the whole of England from 1995 through 2003. The incidence rate for all ages (0–84 years) was 9.21 per 100,000 person-years.

Traditionally, the epidemiology of CNS tumors has been characterized for children and adults of all ages separately in recognition of the differences in pathology and etiology. It is now recognized that the epidemiology of tumors in adolescents is quite distinct from that of older adults,^3,4,18 and hitherto, the incidence of CNS tumors by morphological type has not been described in this age group. The age structure of a population can affect the crude incidence rates. Adjusting to world standard population (as has been done here) allows comparison between registries in different countries because it is independent of the effects of age. The major category of CNS tumors excluded from this analysis is primary CNS lymphoma, which is usually defined as extranodal lymphoma confined to the CNS without evidence of systematic disease.¹⁵ The reliability of this diagnosis depends on the comprehensiveness of staging, and in several studies the diagnosis of primary CNS lymphoma has been revised to systemic non-Hodgkin's lymphoma with possible secondary CNS disease on further investigation.^19–22

Previously published age-standardized incidence rates of primary CNS tumors (based on ICD-O site and not morphological type) in England are 6.5–7.7 per 100,000 person-years in males and 4.5–4.9 per 100,000 person-years in females^1,23 and are lower than incidence rates from this study. These figures significantly underestimated the true burden of CNS tumors because they included only malignant tumors located in the brain²³ or in the brain and spinal cord.¹ They excluded tumors located in the pituitary gland, craniopharyngeal duct, and pineal gland, as well as all nonmalignant tumors.

The reported incidence of CNS tumors in the United States of 14.8 per 100,000 person-years¹⁴ is higher than that given here, while that from Norway, 9.53 per 100,000 person-years,⁸ is similar to ours. The reasons for the higher incidence rate in the United States are three-fold. First, U.S. age-adjusted rates are standardized to the U.S. 2000 population where children younger than 15 years comprised 21.5%, and adults older than 70 years, 9.2% of the total age distribution.¹⁴ These contrast to the world standard population where children younger than 15 years comprise 31%, and adults older than 70 years, 4% of the total age distribution.¹ Standardizing with a population that has relatively older individuals will increase the overall incidence rates because CNS tumors are far more common in these age groups. After adjusting to the world standard population, the overall incidence rate for the United States is 11.61 per 100,000 person-years. Second, there is considerable variation in reporting of nonmalignant (benign and uncertain) tumors in the United States, and the percentage varies from 27% to 60% between different U.S. states.¹⁴ In our data, 31% of CNS tumors were nonmalignant, and if we consider U.S. states with similar percentages of non-malignant tumors (Connecticut, 40% nonmalignant; North Carolina, 38%), their incidence rates (Connecticut, 13.61; North Carolina, 11.43) are well below the overall figure of 14.8 per 100,000 person-years reported by CBTRUS. Finally, we have excluded CNS lymphomas from our analysis, which account for 3% of the incidence in the United States. Thus, one can conclude that the incidence of CNS tumors in the United States is actually not very different from that in the United Kingdom and Europe. These caveats also explain the higher incidence of CNS tumors in females in the United States compared to England and elsewhere,^8,24 which is caused by increased registration of nonmalignant tumors, especially meningiomas (57%), which are more common in females (74%).

The incidence rate of 3.56 per 100,000 person-years seen in children in our study is similar to the incidence rate of 3.3 per 100,000 person-years reported by ACCIS for England and Wales for 1993–1996.²⁵ Corresponding rates for the rest of Europe vary from 4.0 to 5.0 per 100,000 person-years for Iceland, Norway, Finland, and Denmark and between 2.0 and 3.0 per 100,000 person-years for Germany and the Netherlands.²⁵ Data on the incidence of primary CNS tumors for adolescents is scanty. Our previous report showed that the overall incidence rate for malignant CNS tumors in those 15–24 years of age registered in England from 1979 through 1997 is 1.65 per 100,000 person-years.³ The U.S. Surveillance, Epidemiology, and End Results (SEER) program data, which also report only malignant tumors, give an incidence rate of 2.26 per 100,000 person-years for the 15- to 29-year age group.¹⁸ After accounting for the slightly higher rates contributed by the 25- to 29-year age group in those data, this is closer to our rate of 1.94 per 100,000 person-years for malignant CNS tumors in those 15–24 years of age in a more recent time period. Further comparisons by histological group are not possible because the SEER report uses the ICCC rather than WHO 2000 classification. The 2008 CBTRUS report, which uses the WHO 2000 classification, is also unsuitable for comparison because the age groups are different (0–19 years, 20–34 years, and so on).²⁶

With increasing age, not only does the proportion of benign CNS tumors increase but there is also a shift in the spectrum of malignant primary CNS tumors. Pilocytic astrocytoma and embryonal tumors form the bulk of malignant tumors in children. Pilocytic astrocytoma is a WHO grade I astrocytoma, and although regarded as a benign tumor by many,²⁷ it is classified as malignant in the ICD-O first and second editions^28,17 (ICD-O1 and ICD-O2, respectively). However, in the recent ICD-O third edition²⁹ (ICD-O3), pilocytic astrocytoma is classified as uncertain (morphological code 9421/1) rather than malignant in behavior. Future analysis of epidemiology of childhood CNS tumors may show an artificial rise in the proportion of nonmalignant CNS tumors because of this. Embryonal tumors are the second largest group, with 74% medulloblastomas and 26% supratentorial primitive neuroectodermal tumors. Absent among this group are ependymoblastomas (because they share the same morphological code, 9392/3,¹⁷ with anaplastic ependymomas and will have been included there) and atypical teratoid/rhabdoid tumors, which have been recognized as a distinct entity only in ICD-O3.²⁹ Based on a series of cases from single institutions, the incidence of atypical teratoid/rhabdoid tumors is thought to be around 1%–2% of pediatric CNS tumors and at least 10% of all CNS tumors in infants. With increasing use of ICD-O3, incidence data for this group of tumors should be available in the future.³⁰ The incidence of medulloblastoma in childhood from this study is 0.49 per 100,000 person-years and is similar to that reported elsewhere.^15,31 But there is much more variability in the proportion of supratentorial primitive neuroectodermal tumors.^15,32 McNeil et al.³² found that supratentorial primitive neuroectodermal tumors, which had not been described until the 1980s, accounted for up to one-third of all embryonal CNS tumors from 1993 through 1998 in the SEER database. Also seen in infancy are benign tumors of mixed cellular-lineage, such as desmoplastic infantile astrocytoma/ganglioglioma and dysembryoplastic neuroepithelial tumors, which have been traditionally difficult to categorize and are now included in neuronal and mixed neuronal-glial tumors.³⁰ Because these pathological entities have been defined only in ICD-O3, it is not possible to comment on their occurrence in our study.

In adolescents, not only is there a transition from the typical childhood tumor pattern to a distribution more typical of older adults, but there are certain features unique to this age group. First, tumors from the sellar region form a significant proportion (16%) and while in those 0–14 years of age 87% of these tumors are craniopharyngiomas, in those 15–24 years of age 78% are pituitary tumors. Also, this age group has a female preponderance for pituitary tumors, which has been previously reported.^33,34 This is due to prolactinomas and clinically nonfunctioning pituitary adenomas, which are more frequent in females.³⁵ The explanation for this gender-related difference is not clear, but the role of estrogen in tumor promotion and a greater inclination in females to seek medical attention for hypogonadal symptoms has been suggested.³⁵ Second, germ cell tumors peak in incidence in this age group, and germinomas account for the majority of these. They have previously been noted to show a strong male preponderance at the pineal site,³⁶ and this is also seen in our study, where germ cell tumors were eight times more common in males than in females (p < 0.0001).

The data reported here on primary CNS tumors in older adults, including specified diffuse astrocytomas,¹⁴ anaplastic astrocytomas,¹⁴ glioblastoma multiforme,^14,15 oligodendrogliomas,^12,37 anaplastic oligodendrogliomas,¹⁴ mixed gliomas,^14,38 certain ependymal tumors^14,15,39 (myxopapillary ependymomas¹ and subependymomas^15,40), nerve sheath tumors,^{14,15,41–43} meningiomas,^14,15,44 hemangioblastomas,¹⁴ and pituitary tumors,^14,45 are generally similar to data from other studies. There are, however, a few obvious differences, particularly compared with the data from CBTRUS,¹⁴ for reasons discussed above. A final point of note is that astrocytomas NOS and gliomas NOS constituted 18.5% of all primary CNS tumors in our series, and another 13% are of unspecified histology. This is because only 71%–73% of CNS tumors in all ages are microscopically verified.¹ The proportion of tumors with unspecified histology in our series increased with increasing age (Fig. 4), with a median age at diagnosis of 72 years, which suggests that younger individuals are more likely to undergo extensive investigation to achieve specific diagnoses, although this attitude toward the elderly may be changing.⁴⁶ Moreover, it has been shown that, over time, the advances made in neuroimaging, neurosurgery, and neuropathology and improvements in quality of registrations are reducing the incidence of NOS and unspecified tumors.⁴⁷

Little is known about the etiology of primary CNS tumors, and the only proven causes (hereditary syndromes and radiation) account for a small proportion of cases.⁴⁸ The heterogeneous pathologies grouped under CNS tumors further limit our ability to study the etiology of the disease. Analyzing the pattern of these pathologies with age and sex (as has been done here) allows us to speculate about the etiopathogenesis. A nadir in the incidence of CNS tumors at ages 15–19 years and a peak at 75–79 years suggest that both genetic and environmental factors have a role, with the environment being the larger contributor. An increasing incidence of high-grade astrocytoma with increasing age along with a decreasing incidence of low-grade astrocytoma supports the suggestion that malignant transformation of astrocytic cells is a multistep process with sequential acquisition of genetic alterations with age.¹⁵ The peak of CNS embryonal tumors in early childhood and of CNS germ cell tumors in those 15–24 years of age is similar to the incidence pattern seen in non-CNS embryonal tumors (nephroblastoma, neuroblastoma) and gonadal germ cell tumors. This implies that the etiologies for CNS or non-CNS tumors that share the same tissue of origin are likely to be related.

It would also be of interest to analyze the variation in epidemiology by race and ethnicity, as has been done by CBTRUS.²⁶ Currently, the individual-level cancer registration data obtained from the Office of National Statistics is anonymized and does not include information on ancestry/ethnicity. In the future, it may be possible to get such data, and then such an analysis can be done. Moreover, the minority ethnic population in England comprises 7.9%⁴⁹ (4% Asian or Asian British, 2% black or black British, 1.2% mixed, and 0.8% others), which is much less than the 26% in the United States (13.4% black or African-American, 4.4% Asian, 2% mixed, and 6.2% others).⁵⁰ Thus, the incidence rates for England are relatively less likely to be affected by ethnic variation.

Conclusion

In summary, we present a large, comprehensive, and up-to-date analysis of incidence data of primary CNS tumors in the world England that has been obtained from a high-quality national cancer registration system. We have described the epidemiology across the whole of England from 1995 through 2003 for all ages and focused on the changing patterns in children, adolescents, and older adults. The overall incidence is similar to that reported elsewhere in the world but higher than that reported in Britain before. We have also described sex-specific, age-specific, and tumor-behavior–specific standardized incidence rates for all histology groups according to the WHO 2000 classification. We hope this allows other studies to make relevant and meaningful comparisons with our data and that it provides a baseline for secular trend analysis.

R.S.A. is funded by a grant from the Christie Hospital NHS Foundation Trust. R.D.A. and J.M.B. are funded by grants from Cancer Research UK. T.O.B.E. is Teenage Cancer Trust Professor of Teenage and Young Adult Cancer. Parts of these findings were presented as a poster at the 39th Annual Congress of International Society of Pediatric Oncology in Mumbai, October 2007. Data used in this study were contributed by the eight regional cancer registries in England and were provided by Steve Rowan, National Cancer Intelligence Centre, Office for National Statistics, London.

References