Non-malignant meningioma and vestibular schwannoma incidence trends in the United States, 2004–2017

Abstract

Background:

Given concerns about risks associated with growing use of mobile phones over recent decades, we analyzed temporal trends in incidence rates of non-malignant meningioma and vestibular schwannoma in the U.S.

Methods:

Non-malignant meningioma and vestibular schwannoma incidence among adults in the Surveillance Epidemiology and End Results 18 registries during 2004–2017 was evaluated by method of diagnosis: microscopically (MC) or radiographically confirmed (RGC). Annual percent changes (APC) and 95% confidence intervals (CI) were estimated using log-linear models.

Results:

Overall meningioma rates (n=108,043) increased significantly during 2004–2009 (APC 5.4; 95%CI 4.4, 6.4) but subsequently rose at a slower pace through 2017 (APC 1.0; 95%CI 0.6, 1.5). Rates for MC meningiomas changed little from 2004–2017 (APC −0.3; 95%CI −0.7, 0.1), but for RGC rose rapidly until 2009 (APC 9.5; 95%CI 7.8, 11.1) and more modestly thereafter (APC 2.3; 95%CI 1.5, 3.0). Overall vestibular schwannoma rates (n=17,475) were stable (APC 0.4; 95%CI −0.2, 1.0), but MC rates decreased (APC −1.9; 95%CI −2.7, −1.1) whereas RGC rates rose (2006–2017: APC 1.7; 95%CI 0.5, 3.0). For each tumor, the trends by diagnostic method were similar for each sex and each racial/ethnic group, but RGC diagnosis was more likely in older persons and for smaller tumors. Meningioma trends and the proportion RGC varied notably by registry.

Conclusion:

Overall trends obscured differences by diagnostic method in this first large detailed assessment, but the recent stable rates argue against an association with mobile phone use. Variation among registries requires evaluation to improve registration of these non-malignant tumors.

LAY SUMMARY

The etiology of most benign meningiomas and vestibular schwannomas is poorly understood, but concerns have been raised about whether mobile phone use contributes to risk. Descriptive studies examining temporal trends could provide insight but globally, few registries collect these non-malignant cases. In the U.S. reporting became required by law in 2004. We conducted the first large systematic study to quantify and characterize incidence trends for meningioma and vestibular schwannoma according to whether the cases were diagnosed microscopically or only radiographically. Differential trends across registries and by diagnostic method suggest that caution should be used when interpreting the patterns.

PRECIS

Overall trends in meningioma and vestibular schwannoma obscured differences by diagnostic method in this first large detailed assessment. Variation in trends by source of diagnosis among registries and by age at diagnosis and tumor size suggests a potential need to improve ascertainment and registration of these non-malignant tumors, to consider the impact of advances in technology and evaluate demographic and clinical characteristics by method of diagnosis, and to understand these and other features and limitations in interpreting temporal trends for etiologic purposes.

INTRODUCTION

Etiologic factors for most benign meningiomas and vestibular schwannomas (acoustic neuromas) are poorly understood, but concerns have long been raised about whether mobile phone use may initiate or promote occurrence of these central nervous system (CNS) tumors^1,2. Analytic epidemiologic studies assessing mobile phone use and risk of meningiomas have shown no association for duration of use of more than 10 years (Supplementary Table S1). For vestibular schwannomas long-term use has not been consistently linked with risk but there is heterogeneity among investigations, and elevated risks were observed in a few studies for duration of use of more than 10 years (Supplementary Table S2)².

Meningiomas arise from the meninges (the membranes that cover the brain (90%) or spinal cord (10%)). Ninety-percent of meningiomas are considered non-malignant, and the tumors are mostly asymptomatic. Some patients experience headaches that worsen over time, changes in vision, hearing, or loss of smell, problems with memory, seizures, or weakness in arms or legs³. Vestibular schwannomas are benign tumors arising in Schwann cells covering the vestibular cranial nerve leading from the inner ear to the brain and are generally slow growing, but can cause hearing loss, ringing in the ears, dizziness, and problems with balance. Rapidly growing large schwannomas may press against the brain and interfere with vital functions⁴.

Descriptive epidemiologic studies can complement analytic studies to evaluate whether mobile phone use or other changing exposures are linked with brain tumor risk. Multiple studies have evaluated temporal trends in malignant brain tumor incidence^5–7, but few investigations have examined trends in benign brain tumors since reporting of these tumors to population-based registries has not historically been required. Internationally, meningioma rates from the few longer-standing population-based registries to which these benign brain tumors have been reported have been generally stable in more recent years ^8–11, whereas rising rates were reported earlier (e.g., 1980s through early 2000s) in some areas in the U.S.¹², from regional registries ^13,14, from multi-center neurosurgical facilities¹⁵, and among Nordic women⁵. Benign brain tumors have been reportable in the United States starting with cases diagnosed in 2004 (Public Law 107–260; https://www.congress.gov/107/plaws/publ260/PLAW-107publ260.pdf; accessed April 16, 2020). Based on recent data from 18 population-based registries in the U.S. Surveillance Epidemiology and End Results program (SEER-18), Lin and colleagues described rising meningioma incidence overall during 2004–2008 followed by stable rates during 2009–2015¹⁶. With respect to vestibular schwannoma, rates in the U.S. were reported as increasing during 1992–1999¹⁷, but then described as stable during 2004–2010¹⁸. In Nordic countries, there were marked differences in trend patterns by country¹⁹.

An explanation for the mixed findings among studies and in changes in incidence trends may be variations in source of diagnosis. In contemporary practice, the characteristic appearances of meningiomas and of vestibular schwannomas on magnetic resonance imaging (MRI) and/or computed tomography (CT) evaluation permit confidence in radiologic diagnoses^20,21. Historically, the diagnoses of non-malignant CNS tumors (and most types of malignancies) were confirmed by microscopic examination of tumor tissue (pathological review). With the increasing use of and improved resolution of radiographic imaging over time, the proportion of non-malignant CNS tumors radiographically-diagnosed has been increasing. In U.S. studies assessing population-based rates in initially restricted geographic areas during earlier periods to the nationwide assessment since 2004, the proportion of meningiomas and vestibular schwannomas that were radiographically confirmed rose from less than 20% in the 1980s and 1990s^12,17 to more than 50% during the 2000s and 2010s¹⁶,¹⁸. This is mostly due to notable technological advances and expansion in use of radiographic imaging modalities^22,23. Although the recent report by Lin and colleagues (2019) described a large and increasing percent of meningiomas diagnosed by radiographic imaging¹⁶, neither that study nor any other study of temporal trends systematically evaluated and compared temporal trends according to the two methods of diagnosis (microscopically confirmed versus radiographically confirmed) overall or by age at diagnosis, tumor size, or variability within a large geographic region. Using data from the SEER18 registries covering about 28% of the U.S. population²⁴, we have conducted the first large systematic study to quantify and characterize incidence trends for benign meningiomas and vestibular schwannomas according to source of identification (e.g., microscopically confirmed [MC] vs radiographically confirmed [RGC]), patient demographics, tumor characteristics, and registry.

MATERIALS AND METHODS

Data sources

We selected non-malignant brain tumors diagnosed between 2004 and 2017 among persons aged 20 and older residing in the SEER-18 registry areas.

Case definitions

Case definitions used topography, morphology, and behavior codes from the International Classification of Diseases for Oncology, Third Edition^25,26. SEER coding rules allow for topography, morphology, and behavior codes to be assigned to cases even in the absence of microscopic confirmation. These codes are based on the pathology/cytology diagnosis, surgical notes, radiographic diagnosis based on MRI, PET, CT scans, and/or physician statements, as available²⁷.

We identified meningiomas based on a topography code of C70.0 (cerebral meninges), C70.1 (spinal meninges) or C70.9 (meninges, NOS) and morphology codes of 9530–9534 to 9537–9539 (all meningiomas excluding hemangioblastic meningioma 9535). Vestibular schwannomas were those tumors with a topography code of C72.4 (acoustic nerve) and histology code of 9560 (acoustic neuroma). We selected non-malignant tumors with behavior of 0 (benign) or 1 (borderline) and excluded those with a behavior 3 (malignant, N_meningioma=1,261, N_{vestibular schwannoma}=12). Non-malignant brain and other nervous system tumors with the site specified as meninges but the type was not a meningioma (N=754) or site specified as acoustic nerve but the type was not neurilemoma (N=6,484) were excluded. We also excluded tumors with the type specified as meningioma but the site was not meninges (N=920) or type specified as neurilemoma but the site was not the acoustic nerve (N=282). Based on these criteria, we included 108,043 meningiomas and 17,474 vestibular schwannomas.

Variables

We examined incidence rates and trends stratified by method of diagnostic confirmation (microscopic, radiographic, other/unknown). Additionally, we stratified trends by demographic variables: sex (male vs. female), race/ethnicity, and age at diagnosis. Race/ethnicity was categorized as non-Hispanic white, non-Hispanic Black, non-Hispanic Asian/Pacific Islander, non-Hispanic American Indian/Alaska Native (AI/AN) and Hispanic. For age-specific analyses, we grouped age into intervals (20–39, 40–64, 65–79, 80+; and 20–64, 65+). Finally, rates were also stratified by laterality (left, right, other/unknown), tumor size (<2cm, ≥2cm, unknown), and registry (Alaska Native Tumor Registry, Connecticut, Detroit, Atlanta, Rural Georgia, Greater Georgia, San Francisco-Oakland, San Jose-Monterey, Los Angeles, Greater California, Hawaii, Iowa, Kentucky, Louisiana, New Mexico, New Jersey, Seattle-Puget Sound, and Utah).

Statistical methods

Incidence counts and rates were calculated using SEER*Stat (Version 8.3.6), age-adjusted to the 2000 US standard population by 5-year age groups and expressed per 100,000 person-years. Following SEER conventions, only rates based on 16 or more cases are reported²⁸.

The Joinpoint Regression Analysis program (version 4.7.0.0) was used to calculate annual percentage changes (APCs) and 95% confidence intervals (CIs) to quantify trends in incidence. The program selects the best fitting log-linear regression model to identify calendar years when the APCs changed significantly. T-tests were used to determine whether APCs were statistically different from zero. All hypothesis tests were two-sided.

The study was exempt from Institutional Review Board review.

RESULTS

Meningioma

Descriptive findings by source of diagnosis

During 2004 to 2017, 108,043 non-malignant meningioma cases were diagnosed among residents aged 20 years and older in the SEER18 registries, and the overall age-adjusted incidence rate was 12.12 per 100,000 person-years (Table 1). The rate for the cases microscopically confirmed (MC) was 4.97 and for the radiographically confirmed (RGC) was 6.84; the method of diagnosis was unknown for about 3% of meningiomas. Incidence rates among women overall and for each diagnostic method were more than twice those among men. Rates were highest among non-Hispanic Black adults, followed by non-Hispanic whites, with rates lower among non-Hispanic American Indian/Alaska Native, non-Hispanic Asian, and Hispanic adults. Tumor size was not documented for 23% of the MC and 14% of the RGC cases. Overall and among ages 20–64 and 65+ in cases MC, those with smaller tumors (<2cm) had lower rates than those with larger tumors (≥2 cm) whereas among cases RGC, the rate for the smaller tumors was notably higher than the rate for larger tumors. Meningiomas of known laterality (approximately 80%) were equally distributed between right- and left-sided tumors overall and among both diagnostic groups. The RGC rate accounted for 56% of the overall rate with somewhat higher proportions of RGC in females than in males and in white and Black than in Asian/Pacific Islander and Hispanic adults. The RGC proportion increased notably from 33% to 43%, 62%, and 81% among those ages 20–39, 40–64, 65–79, and 80+ at diagnosis, respectively. For cases with larger tumors (≥2 cm), the proportion RGC was 23% among those ages 20–64 versus 57% among those 65+. For cases with smaller tumors (<2cm) the proportion RGC was 75% and 89% in the younger and older age groups, respectively. The RGC rate increased from 50% of the overall during 2004–2008 to 61% during 2014–2017.

Table 1.

Non-malignant meningioma incidence in SEER18, 2004–2017, ages 20+ years.

			Method of diagnosis
	TOTAL		Microscopic		Radiographic
Characteristic	Cases	Rate	Cases	Rate	Cases	Rate	% RGCa
TOTAL	108,043	12.12	44,711	7.97	60,592	6.84	56
Sex
Male	28,117	7.07	13,022	3.10	14,406	3.79	54
Female	79,926	16.49	31,689	6.71	46,186	9.37	57
Race/Ethnicity
NH White	73,603	12.33	28,892	5.00	42,908	7.03	57
NH Black	12,363	14.41	5,165	5.56	6,861	8.43	59
NH AI/AN	652	10.85	288	4.35	335	5.99	55
NH API	8,554	10.35	3,914	4.52	4,399	5.52	53
Hispanic	12,025	10.25	6,088	4.59	5,640	5.37	52
Age
20–39	6,699	2.12	4,367	1.38	2,211	0.70	33
40–64	44,648	10.76	24,279	5.94	19,500	4.62	43
65–79	35,697	33.01	12,962	11.72	21,832	20.43	62
80+	20,999	50.92	3,103	7.69	17,049	41.20	81
Size
<2cm	38,491	4.32	5,638	0.63	31,959	3.59	83
≥ 2cm	49,297	5.52	28,688	3.18	19,881	2.26	41
Unknown size	20,255	2.28	10,385	1.16	8,752	0.99	43
Age<65 by size
<2cm	16,739	2.13	3,682	0.48	12,685	1.60	75
≥ 2cm	24,336	3.14	18,390	2.40	55,734	0.71	23
Unknown size	10,272	1.33	6,574	0.86	3,292	0.42	32
Age 65+ by size
<2cm	21,752	14.48	1,956	1.29	19,272	12.84	89
≥ 2cm	24,961	16.59	10,298	6.83	14,147	9.42	57
Unknown size	9,983	6.66	3,811	2.54	5,460	3.65	55
Laterality
Right	42,776	4.80	16,987	1.89	25,010	2.82	59
Left	43,078	4.83	17,081	1.90	25,153	2.84	59
Other/Unknown	22,189	2.49	10,643	1.18	10,429	1.18	47
Year of diagnosis
2004–2008	30,864	10.60	14,862	5.02	15,274	5.33	50
2009–2013	40,288	12.59	16,276	5.03	22,960	7.22	57
2014–2017	36,891	13.20	13,573	4.86	22,358	8.00	61

^a% RGC: Percent radiographically confirmed based on rates. Incidence rates are age-adjusted to the 2000 US population standard and displayed per 100,000 person-years. NH: non-Hispanic; AI/AN: American Indian/Alaska Native; API: Asian/Pacific Islander

Trends by source of diagnosis

Overall, meningioma incidence rates increased significantly from 2004 to 2009 (APC: 5.4, 95%CI: 4.4, 6.4) and subsequently rose at a slower rate from 2009 to 2017 (APC: 1.0, 95% CI: 0.6, 1.5) (Figure 1, Supplementary Table S3). Rates for the MC cases did not change significantly during 2004–2017 (APC: −0.3, 95%CI: −0.7, 0.1). The RGC rate, however, rose rapidly by 9.5%/year during 2004–2009 (95% CI: 7.8, 11.1) and by 2.3%/year during 2009–2017 (95% CI: 1.5, 3.0). Data regarding tumor size improved dramatically over the study period regardless of diagnostic method (Figure 1, Supplementary Table S3). The rates for meningiomas with unspecified size decreased at −7.4%/year (95% CI: −8.7, −6.2) for the MC cases and at −3.2%/year (95% CI: −4.6, −1.7) for the RGC cases. Rates rose during 2004–2009 for MC cases of both large (≥2 cm) and smaller (<2cm) sizes (APC: 5.1% [95%CI:1.8, 8.5] and 9.0% [95%CI: 3.4, 14.9], respectively), before levelling off during 2009–2017. Rates rose even more rapidly during 2004–2009 for the RGC cases (APC: 8.3% [95%CI: 6.5, 10.1] and 17.0% [95%CI: 14.0, 20.2] for the large and smaller tumors); and continued to increase at a slower pace during 2009–2017 (APC: 1.3% [95%CI: 0.4, 2.1] and 3.9% [95%CI: 2.6, 5.3]).

Figure 1.

Age-adjusted incidence rates of non-malignant meningioma by method of detection and tumor size. SEER18, 2004–2017, ages 20+ years. Incidence rates are age-adjusted to the 2000 US standard population and displayed per 100,000 person-years. Numeric labels indicate annual percent change from Joinpoint analyses and asterisk (*) indicates a statistically significant departure (p<0.05) from slope of 0.

The incidence trends generally were quite similar by method of diagnosis by gender, racial/ethnic group, and age group (Supplementary Figure S1). When stratified by finer size categories, the most rapid increases were for RGC tumors measuring less than 1 cm. The overall trends, method of diagnosis-specific trends, and proportion RGC varied substantially by registry (Supplementary Figure S2). The RGC rates accounted for about half the overall rate in four of the registries, whereas a predominance of RGC rates emerged in five registries and was apparent over the entire period in seven registries.

Vestibular schwannoma

Descriptive findings by source of diagnosis

During 2004–2017, 17,475 vestibular schwannoma cases were diagnosed among residents aged 20 years and older in the SEER18 registries, and the overall age-adjusted incidence rate was 1.92 per 100,000 person-years (Table 2). The MC rate (0.89) was slightly lower than the RGC rate (0.99); the method of diagnosis was unknown for about 2% of the cases. Rates overall and among both diagnostic groups were similar among males and females, highest among non-Hispanic white and lowest among non-Hispanic Black adults. When stratified by age, rates were highest among persons 65–79 overall and for the RGC cases but peaked among those ages 40–64 for the MC cases. Tumor size was not specified for 22% of the MC or 15% of the RGC cases. The rates for larger tumors exceeded those for the smaller among the MC cases, whereas the rates for smaller tumors were 3–4 times those for the larger among the RGC cases. Tumors of known laterality (98%) were equally distributed between the right and left sides overall and among both diagnostic groups. RGC rate proportions were similar among non-Hispanic white, Asian, and Hispanic adults (50–52%), while somewhat higher among non-Hispanic Black and American Indian/Alaska Native adults (60–61%). The proportion RGC rose dramatically with age from 28% among those ages 20–39 to 45%, 72%, and 85% among those 80+ years. The RGC accounted for 30% of the total rate for tumors that were 2.0 cm or larger, in contrast to 67% for the smaller tumors. For cases with larger tumors (≥ 2 cm) the proportion RGC was 22% among those aged 20–64 years to 56% among those 65 and older. For cases with smaller tumors (<2cm) the proportion was 58% and 85% in the younger and older age groups, respectively. The proportion RGC increased from 46% during 2004–2008 to 56% in 2014–2017.

Table 2.

Non-malignant vestibular schwannoma incidence in SEER18, 2004–2017, ages 20+ years.

			Method of diagnosis
	TOTAL		Microscopic		Radiographic
Characteristic	Cases	Rate	Cases	Rate	Cases	Rate	% RGCa
TOTAL	17,475	1.92	8,033	0.89	9,044	0.99	52
Sex
Male	8,343	1.92	3,764	0.86	4,405	1.03	54
Female	9,132	1.93	4,269	0.92	4,639	0.96	50
Race/Ethnicity
NH White	13,016	2.28	5,947	1.09	6,783	1.13	50
NH Black	737	0.78	294	0.30	424	0.47	60
NH AI/AN	110	1.64	44	0.62	65	1.00	61
NH API	1,630	1.84	770	0.85	831	0.96	52
Hispanic	1,663	1.18	878	0.56	745	0.59	50
Age
20–39	2,261	0.71	1,590	0.49	643	0.20	28
40–64	10,186	2.48	5,267	1.31	4,708	1.12	45
65–79	4,226	3.77	1,097	0.95	3,015	2.72	72
80+	802	2.00	79	0.20	678	1.69	85
Size
<2cm	9,236	1.01	2,580	0.31	6,206	0.68	67
≥ 2cm	4,973	0.56	3,441	0.39	1,484	0.17	30
Unknown size	3,266	0.36	1,742	0.19	1,354	0.15	42
Age<65 by size
<2cm	6,249	0.80	2,459	0.32	3,680	0.46	58
≥ 2cm	3,801	0.51	2,929	0.40	843	0.11	22
Unknown size	2,397	0.31	1,469	0.20	828	0.11	35
Age 65+ by size
<2cm	2,987	1.96	391	0.24	2,526	1.67	85
≥ 2cm	1,172	0.77	512	0.33	641	0.43	56
Unknown size	869	0.57	273	0.18	526	0.35	61
Laterality
Right	8,534	0.94	3,902	0.43	4,467	0.49	52
Left	8,599	0.95	3,993	0.44	4,429	0.49	52
Other/Unknown	342	0.04	138	0.02	148	0.02	50
Year of diagnosis
2004–2008	5,635	1.87	2,959	0.97	2,566	0.86	46
2009–2013	6,382	1.94	2,843	0.87	3,381	1.03	53
2014–2017	5,458	1.94	2,231	0.82	3,097	1.08	56

^a% RGC: Percent radiographically confirmed based on rates. Incidence rates are age-adjusted to the 2000 US population standard and displayed per 100,000 person-years. NH: non-Hispanic; AI/AN: American Indian/Alaska Native; API: Asian/Pacific Islander

Trends by source of diagnosis

The incidence of vestibular schwannomas overall was stable during 2004–2017 (APC: 0.4, 95% CI: −0.2, 1.0), whereas the MC rates decreased (APC: −1.9, 95% CI: −2.7, −1.1) and the RGC rates rose (APC_2006–2017: 1.7, 95% CI: 0.5, 3.0), both significantly (Figure 2, Supplementary Table S3). Rates for cases with unknown size declined rapidly at −9.5% (95%CI: −11.3, −7.6) and −7.3% (95% CI: −9.3, −5.3) for the MC and RGC cases, respectively. Rates for large and smaller MC cases did not change significantly over the time period. Rates for larger (≥2 cm) RGC tumors rose at 1.6% (95% CI: 0.1, 3.2) per year during 2004–2017, whereas rates for smaller (<2cm) RGC tumors rose at 16.9% (95% CI: 10.2, 24.0) during 2004–2007 and 5.4% (95% CI: 2.7, 8.2) during 2007–2013 before levelling off during 2013–2017.

Figure 2.

Age-adjusted incidence rates of vestibular schwannoma by method of detection and tumor size. SEER18, 2004–2017, ages 20+ years. Incidence rates are age-adjusted to the 2000 US standard population and displayed per 100,000 person-years. Numeric labels indicate annual percent change from Joinpoint analyses and asterisk (*) indicates a statistically significant departure (p<0.05) from slope of 0.

The trends generally were quite similar by method of diagnosis by gender, racial/ethnic group, and age group (Supplementary Figure S3). The registry-specific trends did not vary as notably for vestibular schwannomas (Supplementary Figure S4) as for the meningiomas. However, the relative frequency of the method of diagnosis did vary, with MC rates predominating in four registries, RGC rates in six registries, and similar rates in the remaining four registries.

DISCUSSION

Based on more than 125,000 cases diagnosed during 2004–2017, we found that the incidence of radiographically-confirmed (RGC) tumors was increasing, whereas incidence of microscopically-confirmed (MC) meningiomas and vestibular schwannomas were stable and decreased respectively. The difference in temporal trends by method of diagnosis was similar for both sexes and among persons of each racial/ethnic group. RGC meningioma and vestibular schwannoma cases were more likely than MC cases to be diagnosed in older people and to be smaller in size at diagnosis. This is the first study to formally evaluate temporal trends according to method of diagnosis and to compare MC and RGC non-malignant meningioma and vestibular schwannoma according to patient and clinical characteristics. We found that the temporal trends in overall incidence rates obscure notably different trends according to method of diagnosis.

Literature Review

Meningiomas

To our knowledge, only one early (1985–1999) U.S. study reported trends by method of diagnosis, identifying meningioma cases (n=5,283) from six population-based U.S. state cancer registries¹². The results showed an overall increase for both MC (Average APC: 1.0, 95%CI: 0.3, 1.7, 4394 cases) and non-MC cases (Average APC: 4.1, 95%CI: 2.5, 5.6, 889 cases)¹². A recent study based on 83,030 meningiomas from U.S. SEER-18 registry data (2004–2015) found that overall rates increased at 4.6% annually during 2004–2009 and then stabilized during 2009–2015, noting, however, significant increases throughout the entire time period for persons less than age 60 years and for tumors ≤3 cm in diameter¹⁶.

Meningiomas and other benign brain tumors have been reported for several decades to Nordic population-based cancer registries. From 1943 to 1982, meningioma incidence in Denmark rose from 0.61 to 2.42 per 100,000 persons 20 years and older (age-standardized to world standard population)²⁹; rates also increased in four Nordic countries (including Denmark) during 1968–1997⁸ and during 1974–2003, particularly among women in the early 1990s⁵. Although the source of case ascertainment in Nordic countries was not reported, lower rates of meningiomas than gliomas suggest it is likely that the rates and trends reflect under-ascertainment of RGC meningiomas. Supporting this hypothesis Christensen (2003) noted a decline from 95% histologic confirmation during 1978–1982 to 79% during 1983–1987 as did Klaeboe (2005) after 1980–1982 for both men and women aged 60–84, consistent with introduction of CT examinations^8,29. In Finland, 19% of the benign brain tumors identified in the Hospital Discharge Registry between 1985 and 1989 were not reported to the Finnish Cancer Registry³⁰.

Outside of the Nordic countries, trends have been mixed. In Israel, rates were stable during 1990–2004 and then declined during 2005–2015 among Jewish men but increased significantly among Arab men and women during 1990–2005 before leveling off during 2006–2015¹⁰. Keinan-Boker and colleagues also noted that an estimated 95% of malignant brain tumors were reported to the national Israeli registry versus an estimated 61% for benign or uncertain histology brain tumors¹⁰. Regional studies in Gironde, France (n=1,321) and Girona, Spain (n=599) reported rising incidence along with increases in non-MC rates that exceeded MC rates for all CNS tumors combined (no information about meningiomas by method of diagnosis)^13,14. Data from 13 pathology laboratories servicing 24 neurosurgical centers in Australia (n=1,865) from 2000 to 2008 showed significant increases in incidence of MC meningiomas in men (APC: 5.3, 95% CI=2.6, 8.1) but not women (APC: 0.6, 95%CI: −3.6, 5.0)¹⁵.

Vestibular schwannoma

One small U.S. study of vestibular schwannoma has evaluated incidence trends according to method of diagnosis. Based on data from up to 11 population-based state cancer registries during 1985–1999, the average APC for MC tumors was 10.2%/year (95%CI: 3.3, 19.2, 161 cases) and for non-MC tumors was 34.8% (95%CI: 19.0, 52.4, 39 cases)¹⁷. More recent data during 2004–2010 on 23,729 patients included in the 98% of the U.S. population covered by the Central Brain Tumor Registry in the United States found stable overall rates (APC: −0.4, 95%CI: −3.4, 2.7) with 49.7% MC and 48.3% non-MC, but trends were not reported separately according to method of diagnosis¹⁸.

In the United Kingdom, overall incidence rose from 2.4 to 7.6 per million during 1980–1997 and subsequently declined to 5.5 per million in 2000, although the total number of cases on which these data were based was not reported³¹. In Denmark (n=2,283), vestibular schwannoma increased from 3.1 per million in 1976 to 22.8 per million in 2004 followed by a decrease to 19.4 per million in 2008³². In four Nordic countries (n=5,133), incidence rose during 1987–2007, but there was significant heterogeneity in rates and trends among the four countries¹⁹. Data for 492 patients with all types of nerve sheath tumors (76% vestibular schwannomas) from 13 pathology laboratories servicing neurosurgical centers in Australia suggested significant decreases (APC: −3.5, 95% CI=−7.2, −0.2) during 2000–2008¹⁵.

Potential Drivers of Observed Trends

The local and global heterogeneity in incidence trends are influenced by changing methods for detection and reporting. This is especially the case for non-malignant brain tumors, which differentiate themselves from other tumors in several ways. Among these are the relatively recent (2004) mandated collection of non-malignant brain tumor records by US cancer registries, the often incidental diagnoses of asymptomatic tumors, and the high proportion of tumors that do not undergo microscopic confirmation. We discuss below how these factors may contribute to the observed trends.

Reporting or case-finding by registries will impact incidence. The heterogeneity apparent at the registry-level supports this statement. Many registries historically have relied primarily on pathology reports for case-finding. Because benign brain tumors may be diagnosed clinically in non-hospital settings and without pathological confirmation, under-reporting is a concern³³. An early 2000s audit of the Finnish Cancer Registry found that only about two-thirds of all cases were in the registry when hospital discharge data and neurosurgery records were additionally queried³⁴.

The rapid increase in cases in our study during the early time period (2004–2009) may be partly attributable to increased case-finding procedures after the introduction of the Benign Brain Tumors Act³³ that required reporting of these tumors. Furthermore, differences in case-finding across SEER cancer registries are likely to explain in large part the broad variation in rates by registry. Registries that receive and review hospital discharge data may have higher meningioma incidence. Furthermore, health care providers are required to report non-malignant brain tumors. However, there is likely substantial variation in adherence with some practices reporting constantly, some inconsistently, and some not reporting at all. A study of US cancer registries belonging to the North American Association of Central Cancer Registries reported that a primary enabler of more aggressive case-finding (e.g., at radiography centers, clinics and physician offices) was financial resources. A primary barrier was a lack of incentive to achieve high completeness of reporting of non-malignant tumors³⁵. At the registry-level and over time, both changing resources and changing practices could produce temporal trends that do not reflect the true incidence in the population. Increases in radiologically-confirmed tumors could be attributable to improved case-finding over time, but the large variation in rates reported by registries suggest that under-reporting of clinically diagnosed cases remains a problem in some regions³³.

Autopsy and natural history studies have confirmed that non-malignant CNS tumors exist latent and undiscovered in the population and that only a subset of these tumors would have likely grown and resulted in symptoms. For example, in a study of 10,033 autopsies during 1950–1982 at a single large hospital, 231 incidental meningiomas (2.3%) were identified; close to 60% were <1 cm in diameter and only about 5% were ≥3 cm³⁶. In a 5-year follow-up of 351 conservatively treated patients with meningiomas identified incidentally, about 63% of these tumors did not grow, 37% had detectable growth, but only 16% of the total became symptomatic³⁷. More recently, a systematic review found that three-quarters of 2,130 RGC-diagnosed meningioma patients undergoing active monitoring did not demonstrate radiological or clinical progression requiring intervention³⁸. An evaluation of all MRIs conducted during 1995–2003 at a large tertiary care medical center led to identification of 8 patients (0.02%) interpreted as having undiagnosed vestibular schwannoma; however detailed evaluation suggested that 3 of the 8 had auditory or vestibular symptoms³⁹. A population-based follow-up study in Olmstead County Minnesota during 1966–2016 revealed 153 cases of vestibular schwannoma, demonstrating incidence increasing from 1.5/100,000 during the first decade to 4.2/100,000 during the last decade; more recent cases had smaller tumors and fewer symptoms and almost 1 in 4 in the last decade was diagnosed incidentally⁴⁰. These studies underscore that the discovery and diagnosis of meningiomas and vestibular schwannomas depends on the intensity of scrutiny. Latent disease that would become symptomatic may be under-diagnosed whereas tumors that would never lead to symptoms may be discovered leading to over-diagnosis. Increasing discovery of asymptomatic, non-malignant tumors would differentially increase the rates of RGC tumors over MC ones as these tumors can be managed with observation initially, sparing the patient an invasive surgical procedure and the potential harms of over-diagnosis²³.

Three potential drivers of changes in detection are access, technology, and clinical practice. The number of MRI machines operating in a geographic region is an example of access. In Norway, a significant association was found at the county-level between the number of MRI scans per capita and the incidence of meningioma⁴¹. For every 250 MRI scans, one additional meningioma was diagnosed⁴¹. Particularly in settings without a national health care system such as the US, access may also vary at the individual level.

Improvements in technology could lead to earlier diagnoses and increases in the absolute number of tumors diagnosed. For example, incidental findings are more likely using high-resolution MRI sequences than standard resolution sequences⁴². As high-resolution MRIs become more common, more incidental diagnoses of benign brain tumors may occur. Beyond detection, technology can affect how tumors are diagnosed and treated. Studies have found that improvements in imaging by magnetic resonance, computed tomography, and cerebral angiography over the last decade have improved the accuracy of meningioma diagnoses providing more prognostic information and better informing surgical management^43–46. This potentially could have led to a perceived decreased need for microscopic confirmation.

Finally, changes in clinical practice could influence the absolute rates, the timing of diagnoses, and the ratio of RGC to MC. Several studies have shown a plateau or downturn in use of imaging in the period around 2009, where we also observe a change in slope in our meningioma incidence trends^47–50. Lower use of imaging could have resulted in fewer incidental diagnoses. It has also been reported that the proportion of patients undergoing surgical treatment decreased in recent decades, even after controlling for tumor size^51,52. This suggests clinical decision-making may have changed, which could contribute to the shift we observed from primarily MC to RGC⁵¹. It has been previously noted that over time, an increasing proportion of non-malignant brain tumors have been RGC¹⁶.

We sought to explore the reasons and implications of method of diagnosis for interpreting temporal incidence trends overall. The SEER registries were an optimal setting to conduct this work as they belong to a unified program with similar practices, guidelines, data collection structures and requirements. The method of diagnosis variable in the SEER files had high levels of completeness (>97%). Furthermore, the inclusion of both vestibular schwannoma and meningioma in our study allowed us to generate hypotheses about how differences in their clinical presentation could explain differences in temporal incidence trends between these two tumor types. A primary limitation of this study was the absence of potentially informative variables from the cancer registry, such as the indication for the imaging that led to radiographic diagnosis. There was a substantial increase in the completeness of the tumor size variable over time. This is a promising improvement in data quality; however, the interpretation of size-specific temporal trends requires assumptions about which cases were missing size and why. Evaluation of a statistical association between mobile phone use and risk of meningioma or vestibular schwannoma was not feasible due to the descriptive epidemiologic study methods employed along with lack of detailed data on history, frequency, and hours per day/week/month of mobile phone use from the benign brain tumor cases and from an appropriate comparison group.

Conclusions

Our observations of increasing RGC tumor rates, particularly for tumors of smaller size and among older adults, and stable MC tumor rates have implications for our etiologic hypotheses and for patients and clinical practice. With respect to etiology, it appears that notable improvements in imaging technology, access, and clinical and registry case-finding practices may be driving incidence trends over time rather than exogenous or host-related risk factors. The potentially large impact of these non-etiologic factors on the incidence trends and the modest increases in meningiomas and stable vestibular schwannoma rates in recent years argue against a substantial effect of the huge increases in mobile phone use on underlying disease risk. This conclusion assumes however, that the latency period is not many decades in length. It is, however, possible that the trends for meningioma in particular may reflect other postulated etiologic factors that the data sources we used have not allowed us to evaluate in detail, e.g., changing trends in menopausal and post-menopausal hormone use by women. In the future, a formal assessment of completeness of non-malignant tumor registration by registry could lead to best practice guidelines for broad adoption. Such best practices likely include automated review of non-pathologic data sources such as radiology records (including free-standing radiology centers) and hospital discharge data repositories.

Within the clinical context, the findings of our investigation highlight increased detection of benign tumors, many of which are likely to have been diagnosed asymptomatically. To date, much of the literature pertaining to incidental detection has been focused on research findings^53,54. Our results support the need for continuing systematic development of benefit vs risk considerations and effective patient-centered best practice management efforts and communication strategies for those who have been incidentally diagnosed with benign tumors^55,56.