Introduction

Primary tumors of the central nervous system are a non-homogeneous group including entities with various tissue origins, different invasive potency and prognosis. Gliomas and meningiomas represent about 75% of these tumors and correspond respectively to 7.8/100,000 and 5.1/100,000 incidence in the Bordeaux area, France and 6.5/100,000 and 6.3/100,000 in United States [1, 2]. During recent decades, the incidence of these tumors has clearly increased in the developed countries [3]. This might be due to better diagnostic strategies but also to changes in exogenous risk factors. Many questions remain about the etiology of these tumors. To date, only two risk factors have been recognized: genetic syndromes (neurofibromatosis type I, Li Fraumeni syndrome) and ionizing radiation [3, 4]. Moreover, some studies have investigated the role of chemical exposures (pesticides, solvents, heavy metals, and nitroso compounds), physical agents (electromagnetic fields, including mobile phones, and head trauma), biological factors (viruses), and immunological conditions (asthma, eczema, auto-immune diseases, and diabetes), but no definitive conclusions can be drawn [3–5].

The role of sex hormones has also been suggested in the outcome of gliomas and meningiomas. This hypothesis is based on clinical, experimental, biological, and epidemiological data. Case reports have shown exacerbation of symptoms due to the progression of meningioma after the placement of a contraceptive implant [6], during the luteal phase of the menstrual cycle [7] or during pregnancy [7–12]. Other studies observed growth of a meningioma in a transsexual patient after estrogen [13] and estrogen-progestin [14] therapies, or development of multiple meningiomas after long-term therapy with a progesterone agonist [15, 16], which regressed after cessation of treatment [16]. These results are in accordance with data from in vitro studies showing a proliferation of meningioma cells when exposed to progesterone and estrogens together [17]. In contrast, growth inhibition and apoptosis induction were observed in glioma cells exposed to estrogens [18–20]. Animal experiments reinforced these data: after transplantation of glioblastoma lines, tumor growth appeared quicker and survival shorter in male rats and mice, compared to females [21–23]. Moreover, ovariectomy in female rats with glioma made their survival comparable to the males. Estrogen administration to these same females slowed down tumor growth, and their survival became again equal to the one of females with intact ovaries [22]. Finally, the role of hormones has also been suggested by the identification of hormonal receptors in tumor tissues [24–29].

From an epidemiological point of view, some data point to the possible role of hormones. For example, the incidence of meningioma is twice as high in women as in men, especially during the reproductive life period [3, 4, 30, 31]. On the other hand, the incidence of glioma is two-fold higher in men [4], this trend starting during adolescence, increasing until 50–54 years old and decreasing after that. This suggests a protective role of female hormones and/or a harmful role of male hormones [32]. Epidemiological studies have also shown an association between meningioma and breast cancer, a pathology for which the role of hormones is now clearly demonstrated [33–35]. Thus, an increased incidence of meningioma was observed in women with a history of breast cancer, while the incidence of breast cancers was increased in women with a history of meningioma, even though it is not possible to conclude definitively about the role of hormones because of other common predispositions to these tumors (sex and age). Genetic polymorphisms could explain this association: indeed, polymorphisms of the BRIP1 gene, which were involved in the repair of BRCA1 gene and associated with breast cancer risk [36], had also been recently associated with meningioma risk [37].

Owing to this background, several epidemiological studies have been conducted in recent years to search for an association between sex hormones and the occurrence of glioma or meningioma. This article reviews the epidemiological literature on the hormonal hypothesis and demonstrates the need for new research developments.

Materials and methods

Article selection

The search was performed on Pubmed and concerned articles until 30 Sept 2010. Concerning hormonal exposure, exogenous sources (intake of hormonal drugs such as contraceptive, hormone replacement therapy, and sterility treatment), and endogenous sources (reproductive events) were considered. Among sex hormones, we took into account, when mentioned, androgens, estrogens, and progestogens (like progesterone). We considered gliomas, meningiomas, lymphomas but we excluded secondary cancers, embryonic tumors (like medulloblastomas), malignant germ cell tumors, craniopharyngiomas, and pituitary tumors because of their direct role in hormonal secretion or inhibition.

Relevant articles in the English or French language were searched with the following request (« hormone exposure » or « hormone replacement therapy » or « pregnancy » or « oral contraceptive » or « ovarian hyperstimulation » or « estrogen » or « progesterone » or « androgen ») and (« brain tumors » or « brain neoplasms » or « astrocytoma » or « glioma » or « brain lymphoma » or « ependymoma » or « meningioma » or « acoustic neuroma ») and « epidemiology » .

Among the 456 articles identified, 47 were selected from the titles and abstract reading. This step excluded articles about children, other types of brain tumors, trophoblastic tumors, and articles on treatment options. Eighteen of the 47 articles reported epidemiological studies about the role of hormones in the occurrence of gliomas and meningiomas [38–55], while the others did not deal with the hormonal hypothesis or were experimental. Three additional articles [56–58] were identified from the references of the 18 above-mentioned articles. Finally, a total number of 21 articles were retained: nine concerning both gliomas and meningiomas [38–40, 47, 50, 51, 53, 54, 58], seven only gliomas [42, 43, 45, 46, 52, 55, 56], and five only meningiomas [41, 44, 48, 49, 57]. As three articles came from the SEARCH program (directed by the International Agency for Research on Cancer), we selected both the main publication by Schlehofer (1999) [50] which included data published by Ryan [58] and the article by Schlehofer (1992) [51], which is more detailed about hormonal parameters. Both articles from the INTERPHONE study were retained because they reported different results (exogenous hormones in Sweden [54] and reproductive history in five northern European countries [53]). Benson published two articles from the Million Women Study: one concerned oral contraceptive use and reproductive history [38] and the other concerned only postmenopausal women and hormonal replacement treatment use [39]. Finally, the review took 20 articles into consideration.

Data collection

The articles were read in order to complete a standardized form enabling the homogenous collection of relevant data across all articles. Details of study design, definition and recruitment of the compared groups (case/control, exposed/non-exposed), tumors studied (histological type, localization, histological, or radiological diagnosis), data collection, and statistical methods were collected.

Exogenous hormonal exposure comprised oral and other contraceptives, hormonal replacement therapy, sterility treatment, and hormones given for other gynecologic problems. When mentioned, formulation and dose of active principles and history of use (duration, years from first or last use, age at first use, present or past use) were collected. Endogenous hormonal exposure consisted in reproductive events: age at menarche, total ovulatory months, periodicity of menstrual cycles, menopausal status (pre/postmenopausal), age at menopause, type of menopause (natural or induced), history of hysterectomy and/or oophorectomy (age at surgery and date of surgery), number of children (defined as number of pregnancies or birth), age at first/last birth or pregnancy, history of miscarriages, medical or other abortions, history of infertility, history of breastfeeding (total duration). Other factors indirectly related to hormonal status were also collected: body mass index (BMI), smoking history, personal and familial history of hormonal tumors (breast, ovary, and uterus), and thyroid diseases. Identification of hormonal receptors (progesterone, estrogens, and androgens) in tumor tissues was also done when mentioned.

Data synthesis

Detailed data on study methods are presented in Table 1. Results concerning exogenous exposure (OC and HRT), age at menarche, age at menopause, menopausal status, pregnancies, breastfeeding, and BMI are presented in the figures, while the other results appear in the text. Strength of associations (odds ratios (OR) and relative risks (RR)) and confidence intervals (CI) from adjusted analysis were selected for presentation. Analysis without next of kin were chosen when both results (with and without next of kin) were given.

Table 1 General characteristics of the studies analyzing the relation between hormonal exposure and brain tumors (meningiomas and/or gliomas)
Full size table

Results

General characteristics of the studies

Table 1 summarizes the characteristics of the 20 selected articles. The first article dated back to 1990 [56] but half of them were published between 2004 and 2010 (n = 11). Most of the studies were case–controls (n = 14), one of them nested in a cohort of Swedish women studying infertility [47]. They were conducted in the general population, except for two hospital-based studies [40, 48]. There were also four large prospective cohorts: one of them concerned two articles. Two cohorts aimed at studying breast cancer: one of nearly 90,000 women in Canada [52] and one of more than one million British women [38, 39]. The third cohort included 121,700 nurses from eleven American states [57] and the last included 225,355 members of the American Association of Retired Persons [55]. Finally, a cross-sectional study was conducted in hospital [41]. Many of the studies were conducted in North America, eleven in the United States, and one in Canada. Others were run in Australia [43] and Northern Europe: Sweden [47, 53, 54], United Kingdom [38, 39, 53] Denmark [53], Finland [53], Norway [53], and Germany [51]. Eight centers in six countries collaborated in the international SEARCH program (Canada, Australia, Sweden, Germany, France, and USA). Most of the studies selected women aged 15–80, but six articles focused on women aged 30–55 years old [57], women over 40 [42, 52], or over 50 [38, 39, 55]. Studies varied in size, including 120–1,657 cases for gliomas, and 60–1,390 cases for meningiomas. Case recruitment was based on cancer registries [38, 39, 42, 44, 45, 47, 49, 52, 55], on the review of medical records [40, 41, 46, 48, 51, 56] or on both of these sources [43, 50, 53, 54]. Definitions of cases varied between studies: most were based on histological criteria [40, 42–48, 50, 51, 53–56]. But in three studies, case definition was based on radiological criteria for all [41] or some [53, 54] of the cases. In the study of Jhawar, definition and recruitment of cases were based on subject self-report [57]. Only six studies specified that tumors were intracranial [40, 44, 46, 51, 53, 54], while one study considered only spinal tumors [49] and six studies both localizations [38, 39, 47, 50, 55, 57]. None of the studies specified whether meningiomas were benign or malignant. In all case–control studies, controls were randomized and for most of them, controls were matched to cases on sex and age and less frequently on place of residence [40, 43, 44, 50, 53, 54, 56] or ethnicity [40, 45]. Two case–control studies used next of kin of cases [56] or spouses of patients hospitalized for back pain [48] without matching. Data were usually collected by face-to-face interviews [40, 43–46, 50, 51, 54], self-administered questionnaires [38–40, 42, 48, 50, 52, 55–57], phone interviews [42, 45, 49, 53, 54, 56], and less frequently by reviewing medical records [41] or crossing databases [47]. Besides age which was always noted, the number and type of potential confounders taken into account varied between studies: educational or socioeconomic level [38–40, 43–45, 53, 54], ethnicity [40, 45, 48, 55], smoking [38, 39, 48, 49, 55], alcohol consumption [38, 39, 49], physical activity [38, 39, 49], history of radiotherapy or high-dose X-rays [48, 49], BMI [38, 39, 57], marital status [40], family history of cancer in first-degree relatives [42], presence of thyroid disorders [48], height [38, 39]. Some analyses were stratified on menopausal status [40, 46, 48] and age at diagnosis [40, 53].

Gliomas

Exogenous female sex hormones

Oral contraceptives (OC)

Nine studies analyzed the role of OC on glioma (Fig. 1). Generally, exposure was assessed by yes/no, and less frequently duration of treatment was considered by classes. Seven of them showed a moderate reduction of glioma risk in OC users, from 7 to 34%, significant in only two studies [40, 45]. This result was consistent with the significant 35% decrease in risk found in another study that analyzed the overall role of hormonal therapies (OC and/or HRT) [50]. Duration of use did not seem to play a great role [38, 40, 46, 52, 54, 55], except in the study of Felini in which the use for 10 years or more was associated with a significant 52% reduction in glioma risk [45]. In one study, current use was more protective than past use (OR = 0.3 (95% CI = 0.1–0.5) versus OR = 0.8 (95% CI = 0.6–1.1)) [45]. Hatch found that the older the use (i.e., time since first/last use of OC), the lower the risk of glioma, although this trend was not statistically significant [40]. Regarding age at first use, no clear trend was found [40].

Fig. 1
figure 1

Glioma risk and oral contraceptive use. Baseline categories included women who had never used OC. * Women who were postmenopausal or not and had ever used hormonal therapy (OC and/or HRT) were compared to women never having used any hormonal therapy (baseline category). † Association was estimated by relative risk

Full size image
Hormonal replacement therapy (HRT)

All studies except one [39] found a moderate lower risk of glioma in women using HRT, taking into account postmenopausal status [39, 40, 45, 52, 54] or not [46, 55] (Fig. 2). The range of this decrease, from 6 to 44%, was comparable to that observed with OC use and was statistically significant in two studies [45, 46]. No link with duration of use [40, 45, 46, 52, 54, 55], age at first use [40], and time since first/last use [40] was observed.

Fig. 2
figure 2

Glioma risk and hormone replacement therapy use. * Postmenopausal women having used HRT were compared to postmenopausal women who had never used HRT (baseline category). † Women who were postmenopausal or not and had used HRT were compared to women who had never used HRT (baseline category). ‡ Women who were postmenopausal or not and had used hormonal therapies (OC and/or HRT) were compared to women who had never used any of them (baseline category). § Association was estimated by relative risk

Full size image

A more protective effect for current HRT use compared to past use was found in one study (OR = 0.4 (95% CI = 0.2–0.6) versus OR = 0.8 (95% CI = 0.5–1.2)) [45], while no difference between past and current use was found in two other studies [39, 55].

One study measured glioma risk according to type of HRT (estrogen/progestin) and no association was found, even taking into account duration of estrogen or progestin use [55]. In a recent article [39], current estrogen-only HRT users (versus never users) presented a significantly increased risk of glioma (RR = 1.3 (95% CI = 1.05–1.7)), which was stronger for less than 5 years of use (RR = 1.8 (95% CI = 1.2–2.7)). In the same article, no difference in risk was observed according to the estrogen or progestagen compound, the route of estrogen-only HRT (oral/transdermal) and the regimen of progestagen (sequential/continuous).

Furthermore, interaction between OC and HRT use was analyzed in postmenopausal women [45]. Women having used HRT had a lower glioma risk which was unchanged whether they had used OC or not.

Other hormonal treatments

One study reported no association between other contraceptive methods (intrauterine device, subcutaneous implant, intramuscular injection) and glioma, even when regarding duration of exposure [54]. A lower risk in women having taken hormones for gynecologic problems (bleeding, irregular menstruation) was found (OR = 0.4 (95% CI = 0.2–1.1)) but it was not significant, perhaps due to small sample size [54].

Reproductive life

Thirteen studies analyzed the role of endogenous hormones on glioma risk [38, 40, 42, 43, 45–47, 50–53, 55, 56].

Age at menarche

An increased risk of glioma was showed in five studies when menarche was delayed (Fig. 3). Indeed, after 14 years, the increase in risk ranged from 1.4 to 2.1. This increase in risk was found in earlier ages in two studies: for the 13–14 years group [40] and for the 12–13 years group [52]. However, contradictory results were found in the study of Huang for premenopausal women [46] and in the study of Wigertz for women over 50 years [53].

Fig. 3
figure 3

Glioma risk according to age at menarche (years). * No OR was calculated for women with age at menarche less than 11 years. † Association was estimated by relative risk

Full size image
Role of menopause

Eight studies dealt with menopausal status and glioma risk and results appeared controversial (Fig. 4). Only one study showed a significantly increased risk of 62% in postmenopausal women [50], five studies a non-significant increased risk between 10 and 77% [45, 51–53, 55] and the latter two a non-significant decreased risk from 10 to 15% [40, 43]. No difference in risk was observed according to the type of menopause [40, 45, 46, 50, 51, 55]. Likewise, no association was found with history of oophorectomy, age at oophorectomy or time since oophorectomy [40, 51, 55]. Finally, no association or trend was observed with age at menopause [40, 43, 45, 46, 53, 55].

Fig. 4
figure 4

Glioma risk and menopausal status. Baseline categories included premenopausal women. *Association was estimated by relative risk

Full size image
Role of pregnancy and breastfeeding

When studying the influence of pregnancy on glioma risk, the number of self-reported pregnancies at different stages [38, 40, 43, 45, 51–53] and/or the number of births [40, 42, 45–47, 53, 55] were considered. In most of these studies, glioma risk was not influenced by these factors (Fig. 5) or by age at first pregnancy/birth [38, 40, 42, 45–47, 52, 53, 55]. Nevertheless, two studies found a statistically significant decreased risk for women with at least one live birth compared to women without children [42, 47]. Trend with the number of children was unclear, but two studies showed a significant decreased risk for women who had five children or more: Hatch [40] (OR = 0.45 (95% CI = 0.21–0.99)) and Lambe [47] (OR = 0.6 (95% CI = 0.4–0.9)).

Fig. 5
figure 5

Glioma risk in women with at least one pregnancy or live birth. * Women who had at least one live birth were compared to nulliparous women (baseline category). †Glioma risk was estimated by relative risk. ‡Women who had ever been pregnant were compared to women who had never been pregnant (baseline category). §Women having had at least one full-term pregnancy were compared to women having had none (baseline category). || Women with at least three pregnancies of 4 months or more were compared to women with less than three pregnancies of more than 4 months

Full size image

The possible effect of breastfeeding on glioma risk was also analyzed in three studies (Fig. 6). No significant association was found. But longer cumulative lifetime duration of breastfeeding was associated with an increasing risk, statistically significant in two studies [46, 53].

Fig. 6
figure 6

Glioma risk and breastfeeding. *Baseline categories included women who had never breastfed. †Only parous women

Full size image
Other hormonal factors

Other factors possibly affecting hormonal exposure in women had been analyzed in a few studies. No clear relation with glioma was found for cumulative lifetime number of menstrual cycles [45, 46], menstrual cycle regularity [46], history of miscarriages, medical or other abortions [46], hysterectomy [43, 55], or BMI [38].

Some authors considered the effect of hormonal exposure by different histological subtypes of glioma [39, 42, 45, 46, 55]. Huang, Kabat and Benson found no difference with results observed when considering all glioma subtypes together [39, 46, 55], whereas Felini found that the increased risk for an age at menarche higher than 14 years was significant and stronger for gliomas other than glioblastomas [45]. Cantor noted a decrease in risk of glioblastomas stronger than astrocytomas in multiparous women [42].

Meningiomas

Exogenous female sex hormones

Oral contraceptives (OC)

Seven studies were interested in the potential role of OC in the occurrence of meningiomas. Results were not consistent across studies (Fig. 7). Three of them found no association between meningioma risk and OC use, even when taking into account duration of use [38, 40, 54]. A significant reduced risk with OC use was found in the studies of Lee [48] and Preston-Martin for more than 3 years of OC use [49]. On the other hand, a non-significant increased risk was observed in one study, but not with duration of use [44]. As seen in Fig. 7, results of Jhawar were divergent between current and past users [57], which were not found in two other studies analyzing past/current use. Indeed, one study observed a non-significant increased risk [44], whereas the other one found a significant decreased risk [48] in both user groups. No association with time since last or first OC use [40, 44] or age at first use [40, 48] was found.

Fig. 7
figure 7

Meningioma risk and oral contraceptive use. Baseline categories included women who had never used OC. *Association was estimated by relative risk

Full size image
Hormonal replacement therapy (HRT)

Seven studies analyzed the effect of HRT in postmenopausal women [39, 40, 44, 48, 49, 54, 57] and the study of Blistheyn in women aged 26–86 years, irrespective of their menopausal status [41]. Results of these eight studies were heterogeneous (Fig. 8). A marked significant increase in meningioma risk for HRT users was found in two studies [41, 54] and in two other studies for current HRT users only [39, 57]. In the other studies, HRT use had no net effect [44] or tended to have a slightly protective effect [48, 49, 56]. The study of Blistheyn observed a risk of about two for ages beyond 55 years, and a risk of four before this age, in women who could have possibly taken HRT for other reasons than menopause [41]. No clear trend with duration of use was found [40, 44, 48, 49, 54]. Results concerning past/current use were contradictory [39, 44, 49, 57]. Two studies found a reduction [49] or an increase of risk [39] in both user groups, stronger and significant for current users. In the other studies, a significant increased risk in current users only [57] and a non-significant decreased risk in past users only [44]. One study showed that first use before 40 years old was protective regarding meningioma risk (OR = 0.5 (95% CI = 0.2–1.2)) and first use beyond 50 years old tended to increase the risk (OR = 1.2 (95% CI = 0.6–2.4)) [40].

Fig. 8
figure 8

Meningioma risk and hormone replacement therapy use. Referent groups included postmenopausal women who had never used HRT, except for the study of Blistheyn which defined women who were postmenopausal or not, aged from 26 to 86 years and had never used HRT as baseline category. *Association was estimated by relative risk

Full size image

The relation between meningioma and type of HRT was examined in two studies. One found that estrogen-only HRT use had no effect on meningioma risk (OR = 0.9 (95% CI = 0.5–1.6)) while combined HRT use tended to increase this risk by 30% (OR = 1.3 (95% CI = 0.6–2.8)) [44]. The second article found a significant increased risk with current estrogen-only HRT use (RR = 1.4 (95% CI = 1.03–2.0)), which was more pronounced with estradiol (RR = 1.9 (95% CI = 1.3–2.8)) especially when it was administered orally (RR = 2.4 (95% CI = 1.3–4.5)) [39]. No association was found with estrogen–progestagen HRT, even taking into account the compound or regimen of progestagen.

Other hormonal treatments

In one study, the use of other contraceptive methods (intrauterine device, subcutaneous implant, intramuscular injection) was associated with an increased risk of 1.50 (95% CI = 0.90–2.60), reaching 2.50 (95% CI = 1.00–6.30) for a duration of use higher than 5 years in one study [54]. Regarding hormonal treatments for gynaecologic problems (bleeding, irregular menstruation), no clear relation with meningioma was observed, even with the duration of use [54].

Reproductive life

Age at menarche

Six studies examined the effect of age at menarche on meningioma risk (Fig. 9) and only one study found a strong but non-significant increase in risk with age at menarche [57]. No trend was observed with increasing age at menarche in other studies.

Fig. 9
figure 9

Meningioma risk according to age at menarche (years). *Association was estimated by relative risk

Full size image
Role of menopause

The effect of menopause was examined by eight studies and appeared clearly in several of them (Fig. 10). The increased risk was non-significant and moderate in two studies [40, 50], while it was significant and ranged from two to five in three studies [44, 48, 49]. In contrast, a risk reduction of 48–60% was observed in postmenopausal women in two studies [51, 57], significant in one of them [57]. No association was found in the study of Wigertz [53]. Moreover, in some studies, the effect of menopause was studied by taking into account type of menopause. Hatch found an elevated 13% global risk in postmenopausal women, but this increase disappeared in the surgical menopausal group (OR = 1.0 (95% CI = 0.5–1.8)), whereas it was stronger but non-significant in the natural menopausal group (OR = 1.5 (95% CI = 0.7–3.1)) [40]. The decreased risk among women with bilateral oophorectomy reinforced these results [40]. Similarly, Schlehofer noted a greater reduced risk in surgical menopausal women (OR = 0.12 (95% CI = 0.01–1.30)) than in natural menopausal women (OR = 0.59 (95% CI = 0.18–1.94)) [51]. This risk difference by type of menopause was not observed in two other studies [48, 51]. Regarding age at menopause, no clear trend was seen [40, 49, 53].

Fig. 10
figure 10

Meningioma risk and menopausal status. Baseline categories included premenopausal women. * Women who were postmenopausal and had never used HRT were compared to premenopausal women. † OR or RR were recalculated from those presented in the articles, by taking premenopausal women as referent groups. ‡ Meningioma risk was estimated by relative risk

Full size image
Role of pregnancy and breastfeeding

Results of nine studies analyzing the effect of pregnancy (at different stages) [38, 40, 48, 49, 51, 53, 57] or live births [40, 44, 47, 53] were not consistent (Fig. 11). Several found a slight increased risk of about 20–30% in women with children [38, 40, 51, 53], and up to a two-fold risk in the studies of Jhawar [57] and Custer [44]. However, these associations were not significant. In contrast, two studies showed a reduction of risk with pregnancy [48, 49], significant in one study [48], and Lambe found no clear difference between women with or without children [47]. No distinct trend with the number of pregnancies or births was found [38, 44, 57], except for one study which observed a trend to a reduction of risk with the number of pregnancies [48]. Moreover, the influence of age at first pregnancy/child was not obvious [38, 40, 44, 47–49, 57].

Fig. 11
figure 11

Meningioma in women with at least one pregnancy or live birth. * Women who had ever been pregnant were compared to women who had never been pregnant (baseline category). † Women who had at least one live birth were compared to nulliparous women (baseline category). ‡ Women having had at least one full-term pregnancy were compared to women having had none (baseline category). § Women with at least three pregnancies of 6 months or more were compared to women with less than three pregnancies of 6 months or more. || Association was estimated by relative risk

Full size image

The role of breastfeeding on meningioma risk was analyzed in two studies (Fig. 12). A reduced risk was observed ranging from 16 to 30%. This reduction was more important for an overall duration of breastfeeding higher than 17 months [40], but non-significant.

Fig. 12
figure 12

Meningioma risk and breastfeeding. Analysis concerned only parous women. *Baseline categories included women who had never breastfed

Full size image
Others hormonal factors

The influence of BMI was considered in four studies. Except for one study which found a moderately reduced risk with overweight/obesity [48], an increased risk was observed with overweight [57] or obesity [38, 44], significant in one study (RR = 1.40 (95% CI = 1.08–1.87)) [38]. Other factors potentially modifying hormonal exposure (thyroid diseases, female hormone-related cancer, and smoking) were studied by Lee [48]. The risk was significantly doubled in women with hyperthyroidism, Hashimoto’s disease, benign, or malignant thyroid tumor (OR = 2.0 (95% CI = 1.8–4.9)). There was a non-significant reduced risk in women who had a history of female hormone-related cancer (breast, endometrial, or ovarian cancer) (OR = 0.6 (95% CI = 0.2–1.6)), which was similar when only breast cancer was considered (OR = 0.8 (95% CI = 0.2–2.6)).

Hormonal receptors

The hormonal receptor expression was taken into account in one study [44]. As estrogen receptor was expressed in a small number of meningiomas, analysis concerned only tumors expressing progesterone receptors (PR). Thus, a significant association of OR = 3.2 (95% CI = (1.3–8.0)) was observed between OC use and risk of tumor expressing low PR (in 0–25% of cells). Regarding the influence of pregnancy, women with at least one child had a highly increased risk of meningioma expressing low PR (OR = 4.0 (95% CI = 0.7–22.3) for one birth, OR = 4.3 (95% CI = 1.0–18.8) for two births and more). This pattern of results was not found for meningiomas expressing high PR. No obvious association was found between tumors expressing low or high PR and menopause, HRT, or BMI.

The body of results for each type of brain tumor has been summarized in Fig. 13. However, some studies were not included for several reasons: no overall risk estimated, divergent trend of risk among subgroup or trend of risk toward the null risk (OR = 1.00 or RR = 1.00).

Fig. 13
figure 13

Summary of results on the association between main hormonal factors and risk of brain tumors. n number of different articles examining the parameter. Birth corresponded to live birth and full-term pregnancy. Pregnancy lasting 4/6 months or more was included in “pregnancy”. *Studies with no overall risk or divergent risks by subgroup were excluded. †Study with OR or RR = 1.00 were excluded since no conclusion can be drawn. ‡For two studies, significant increased risk concerned only current HRT use

Full size image

Discussion

This review identified a limited number of epidemiological studies mainly published in the last decade on the role of hormones in the occurrence of brain tumors (n = 20). The most consistent results were an apparent decrease in glioma risk in women treated by HRT and CO, an increase in glioma risk for women with late menarche or menopausal women, and an increase in risk of meningioma with HRT and menopause. Other hormonal factors did not show a clear association with meningioma and glioma.

Clinical, biological, and epidemiological data detailed in the introduction suggested that female sex hormones may promote meningioma and inhibit glioma. Among the main results, some of them are consistent with biological hypothesis which assume female hormones have a protective effect against glioma. First, older age at menarche was significantly associated with glioma. This result is of particular interest for biological reasons. Older age at menarche implies a lower number of menstrual cycles between menarche and menopause and thereby to a shorter cumulative time of estrogen exposure. Furthermore, some authors reported a long-lasting increase in free estrogen and androgen levels and/or a decrease in sex hormone-binding globulin (SHBG) levels in women having had an early menarche [59–63]. Cellular and animal studies showed that estradiol directly affected glioma cells [18–20] and could inhibit glioma growth [18–20] through several mechanisms (inhibition of cell cycle entry [64], lower glutamate toxicity in glial cells [65], down-regulation of astrocyte-derived TGF-β1 [66], and melatonin-driven neuroprotection via aromatase [67]). Secondly, although the association between menopause and increased glioma risk is not clear and significant, it is consistent with the biological hypothesis. However, further studies are necessary to provide additional results. Finally, HRT use seems to lower the risk of developing a glioma, this finding being reinforced by the risk reduction observed with OC use. Nevertheless, this association should be interpreted with caution since no clear trend with duration of use, age at first use, or time since use has been observed. As HRT contains estradiol alone or in combination with progestins, physiological mechanisms related to estradiol and described above could also be involved. On the other hand, the effects of progestins on glial cells are unclear. Some evidences for progesterone neuroprotection have been demonstrated, such as regulation of myelin formation through progesterone receptors (PR) by activating transcription [68], but further investigations are required to clarify the role of progesterone on glial cells. Gliomas expressed hormonal receptors which in descending order of frequency were androgen, progesterone, and estrogen receptors [24]. The rarity of estrogen receptors could limit the influence of estrogen on glioma development. However, some estrogen effects may not be mediated by receptors [69], and estrogen can also regulate PR expression in the brain [70, 71]. The presence of two isoforms of PR (PR-A and PR-B) might increase with the grade of astrocytic tumors [24]. Their functions in glial cells and in gliomagenesis are unknown, although some data suggested that progesterone may promote the growth of tumor cells via PR-B while inhibiting the growth via PR-A [72]. These observations require further investigation, since normal cerebral tissue and numerous tumors of varying malignancy also express these receptors [73].

Regarding meningioma, an increase in risk was found with HRT, consistent with biological hypothesis. This result was not found in all studies, but it was observed in the two prospective cohorts, in which bias are expected to be less important. The striking increase in risk seen only in current users could be the result of an immediate and temporary effect of hormones (especially estradiol) on tumor growth, decreasing after cessation of exposure. However, these results should be confirmed in further studies, since no relation with duration of use was observed. An increased risk was also found with menopause, which is conflicting with hypothesis suggesting a promote effect of female sex hormones on meningioma. This could be explained because most studies did not take into account HRT use. One study [57], which analyzed the effect of menopause in HRT non-users, found a significant reduced risk, consistent with the biological hypothesis. Thus, it seems difficult here to specify the effect of menopause on the occurrence of meningioma independently of HRT use.

The weakness of the epidemiologic evidence may reflect possible bias and differences between studies in terms of definition of hormonal exposure. With selection bias, exposure misclassification is the main source of bias because all studies except one used self-reported information. However, the importance of this problem varies according to hormonal factors and concerns mainly retrospective studies. Indeed, remembering age at menarche/menopause, number of pregnancies, and their dates seem easy, while it is more difficult to precisely report the lifetime hormone therapy use history. Thus, only three studies (mainly prospective studies) have analyzed the type of treatment preparation because of missing data. This non-differential exposure misclassification leads to underestimate the risk. Differential misclassification is also possible due to the fact that on one hand, cases might recall exposures better than controls and on the other hand, their memory might be altered because of the tumor (especially for severe cases such as gliomas). Possibility of misclassification is also greater with next of kin, which represents from 1 to 83% of subjects [42]. However, results were similar when comparing analysis with and without next of kin. Concerning prospective studies, which enable precise and reliable data collection, some studies did not always collect lifetime hormonal treatment use [52, 57] and weight changes during life [38].

Heterogeneous results could also be related to the fact that parameters calculated from standard questions on hormone use and reproductive events do not reflect actual exposure to circulating hormones at different times throughout women’s life. Furthermore, different definitions of hormonal exposure were observed between studies. For example, pregnancies have been considered at very different stages. Progesterone and estrogen levels increase during pregnancy, especially in the second and third trimesters. So, contrary to number of live births, results for number of pregnancies might have been diluted by early pregnancy loss. Definition of menopause either included or not the following criteria: surgical menopause, a reference age between 45 and 55 years, presence of amenorrhea, duration of amenorrhea, HRT use. This has led to different groups being compared. Another point is that some authors did not take into account exposure occurring shortly before the diagnosis, since it was assumed that it was not likely to play a role in the disease. This delay ranged from 1 or 2 years [40, 43, 46, 47] to 10 years [44, 48]. In parallel, Jhawar took into account a minimal length of exposure, for instance 2 years of OC use [57].

Methods and results are also heterogeneous because of different medical practices and socio-cultural habits regarding OC and HRT use in different countries and at different times. In Asia, Latin America and in Anglo-Saxon countries, the most widely used birth control method is sterilization. On the contrary, sterilization is not common in Europe, where the pill is the main contraceptive method. Moreover, formulations, compositions, and dosages have changed considerably since the late 1960s, with appearance of mini- and even micro-pills and second- and third-generation progestins. Same temporal and spatial differences exist for HRT.

Regarding interpretation of results, it can be difficult to identify effects of each hormonal factor. For example, OC use is often associated with ovulation inhibition and thereby modification of endogenous hormone secretion. Similarly, during breastfeeding, levels of specific hormones (prolactin, oxytocin) raise while normal ovulation is stopped. It is also difficult to analyze breastfeeding effect separately from other aspects of pregnancy. Furthermore, it seems that long-term hormonal changes occur after a first pregnancy and that subsequent pregnancies do not confer additional protection. Indeed, parous women have higher levels of SHBG and lower levels of androgen, prolactin and free estradiol than nulliparous women [74–77]. So, it is noteworthy that analysis of number of pregnancies/births is not necessarily relevant and that early loss pregnancy, possibly unnoticed by women, could also confer a protection.

Despite discrepancies between studies, lack of power and low levels of risk measured, hormonal factors concern all women for endogenous exposure and most of them for exogenous exposure. Moreover, prognosis of many brain tumors is particularly poor. Therefore, there is a need to conduct new studies taking into account the limitations of existing data. One possible way forward is to improve exposure assessment to hormonal treatments, e.g. specifying exogenous amounts of hormones by considering the composition and dosage of contraceptives and HRT, and their route of administration. Cumulative exposure scores by type of hormone might be calculated from the combination of dosage and duration of use and used to search for a dose–effect relationship. Other hormonal treatments like those for infertility [78] or for other reasons (gynaecologic or not) should be analyzed. The role of menopause should be examined by taking into account HRT use. To date, most studies have not included all different exposures in a multivariate analysis. This analysis could help to better assess effect of each hormonal factor, but could also be complicated due to possible correlations between factors. It would also be pertinent to clarify the influence of histological subtype, tumor location, and hormonal receptor subtypes. As fat tissues are a reservoir for steroid hormones, obesity and excess weight influence SHBG concentration and circulating hormone levels. Other hormonal disturbances affecting follicle stimulating hormone, luteinizing hormone, or thyroid-stimulating hormone could also play a role in the development of glioma and/or meningioma. Thus, it would be interesting to precise the potential role of BMI and medical history, such as thyroid diseases and other potentially hormone-dependent cancers. Finally, the role of other hormones such as androgens should be further analyzed in biological and epidemiological studies. Indeed, androgen levels vary during women’s life and in special events like surgical menopause (fall of androgen level) and earlier menarche (rise of androgen level).

This literature-based survey on the role of sex hormones in the occurrence of meningioma and glioma follows on from the review of Claus published in 2007 [79], which included seven studies until 2006. That review has examined effect of hormonal treatments (OC and HRT) on meningioma risk and has given recommendations for managing women with meningioma (i.e., information on the benefits and risks of these treatments, and role of hormonal receptor analysis for prognostic and therapeutic purposes). Our study now goes further by examining also endogenous hormonal factors and glioma studies and highlights the need to further explore the role of hormonal factors since no definitive association can be drawn. To date, precaution on hormone treatment use or other modifiable factors (i.e., induced menopause) cannot be recommended in management of women with meningioma or glioma. Furthermore, several studies examined the effect of antihormonal therapy (antiprogesterone and antiestrogen agents) especially on meningioma [80–86] and showed no effect, which suggests that the relation between sex hormones and brain tumors could be more complex.