To our knowledge, this is the largest cohort (n=1,292,462 women) published to date to examine the relationship between reproductive factors (parity and age at first birth) and brain cancer risk. In this prospective cohort study, we found a statistically significant decreased risk of brain cancer with increasing parity. Our findings of a reduced risk of brain cancer associated with higher parity is in agreement with previous studies
], but is not in agreement with other studies that reported no association with parity
The mechanism by which increased parity may confer protection against the future development of brain cancer in women remains unclear. Pregnancy elevates serum estrogen levels approximately 100-fold
]. Increasing parity is associated with an overall increase in lifetime estrogen exposure. There is experimental evidence that estrogen inhibits proliferation of glioma
] and induces cell death
]. In addition, estrogen has been shown to reduce glutamate toxicity in glial cells
] and speeds up the process of repair after brain trauma
]. Thus, if estrogens are associated with a reduced risk of brain cancer, we would expect pregnancy to offer some protection from brain cancer. Our data provide support for this hypothesis. However, it is unknown whether estrogen or other hormones play a role in the development of brain cancer. Support for a possible role of sex hormones in gliogenesis comes from the fact that some gliomas express estrogen receptors
]. Furthermore, aromatase, an enzyme with high levels in human and rat glioblastomas, can convert testosterone to estradiol
In the present prospective cohort study, we found that risk of brain cancer increased with increasing age at first birth after adjusting for parity. Our finding of a positive association between age at first birth and brain cancer risk agrees with one previous study
] but not with other studies that reported no association with age at first birth
]. The reasons are unknown. It may be that a younger age at first birth played a protective role in brain cancer risk via elevated levels of some hormones (including estrogen and progesterone) during pregnancy. Moreover, it has been found that age at menarche (and thus exposure to period estrogens stimulation) is related to age at which a women delivers her first child
]. Earlier exposure to regular menstrual cycles and the concomitant earlier increase in exposure to estrogens may protect against tumor initiation during this early life stage
]. Our finding of an increased risk of brain cancer associated with older age at first birth is in keeping with the hypothesis that estrogen exposure is protective with respect to brain cancer risk. However, because there is only limited evidence to date for a positive association between age at first birth and risk of death from brain cancer, the possibility that this is a chance finding also needs to be considered. Clearly, more work will be needed before the influence of age at first birth on the risk of brain cancer is understood.
To our knowledge, ours is the first prospective cohort study to report a protective effect of parity and early age at first birth on the subsequent risk of death from brain cancer. However, these results are inconsistent with most studies published to date. The reasons for these differences in findings are unknown. Only three prospective studies have examined the relationship between parity and age at first birth and brain cancer risk
]. Most of previous studies used case–control design
]. The main strength of this investigation is its prospective study design, which eliminates the possibility of recall bias. Also, the complete population coverage and follow-up made possible by the national identification number has left the study without selection bias. The mean age at baseline was 24.33 years in this study. Women included in our study tended to be younger than those in previous cohort studies (ranging from 48.5 to 55.9
]. In addition, the number and type of potential confounders taken into account varied between studies. The relative risk estimate may therefore be different between studies
]. Also, information on exogenous hormone use (the use of oral contraceptives (OC) and hormone replacement therapy (HRT)) and cumulative number of menstrual cycles (age at menarch and menopause) were not available in this study. Previous prospective cohort studies have included all different exposures in a multivariate analysis. This analysis could help to better assess effect of each hormonal factor. Finally, age at first birth may be different in different countries and during different periods. The mean age at first birth was 24.33 in our cohort. This value was increased during the study period (ranged from 23.3 to 28.1 in 2006). Likewise, same temporal differences may exist for European countries.
Mortality data have been widely used to generate epidemiologic hypotheses, despite their inherent limitations. The completeness and accuracy of the death registration system should be evaluated before any conclusion based on the mortality analysis is made. In the event of a death in Taiwan, the decedent’s family is required to obtain a death certificate from the hospital or local community clinic, which then must be submitted to the household registration office in order to cancel the decedent’s household registration. The death certificate is required in order to have the decedent’s body buried or cremated. Death certificates must be completed by physicians in Taiwan. It is also mandatory to register all deaths at local household registration offices, the death registration is accurate, reliable and complete.
Taiwan is a small island with a convenient communication network. It is believed that all brain cancer cases had access to medical care. Mortality data rather than data on inpatient cases was used to assess the association between parity, age at first birth, and brain cancer in this study. The mortality of a disease is a function of its incidence and fatality. The 5-year survival rate for brain cancer has been reported to be as low as 30% in United States
]. Deaths from brain cancer may therefore be regarded as a reasonable indicator of the incidence of brain cancer.
Several studies have found that the use of OCs was associated with a reduced risk of brain cancer
]. Other studies, however, have failed to find statistically significant association between the use of OCs and brain cancer
]. Regarding HRT, previous studies have largely reported no association between HRT and brain cancer risk
]. We were unable to adjust for these two hormonal factors in the current study due to the lack of available data. Since the use of OC and HRT are low in Taiwan compared with Western countries
], the confounding effect resulting from these two factors should be small, if any exists at all. Furthermore, if the association between these two potential confounding variables and the risk of brain cancer is not as strong as the one that has been observed for parity and age at first birth, adjustment of these variables will not qualitatively change the conclusion.
Smoking was associated with the incidence of brain tumors
]. There is unfortunately no information available on individual smoking habits and thus it could not be adjusted in the analysis. The smoking prevalence for females was only about 4.2% in Taiwan
]. We think that the degree to which not controlling for this variable may have affected our results should be small if it existed at all. Nevertheless, the problem of possible confounding from smoking should be evaluated. Smoking habits account for at least part of the socioeconomic differentials, with rates of smoking considerably higher among those with lower levels of education. In this study, years of schooling was used as a proxy for socioeconomic status and was included as a control variable in the multivariate analysis. We therefore may have partially indirectly adjusted for the confounding effect resulting from smoking.
Ionizing radiation, occupational exposure to solvents, and electromagnetic fields are the most important risk factors for brain cancer
]. There is no information available on these variables for individual study subjects and thus they could not be adjusted for in the analysis. However, there is no reason to believe that there would be any correlation between these variables and parity and age at first birth.
Some potential limitations of this study needed to be noted. First, our computerized death-certificate database does not specify the tumor type and the histological classification for the brain cancer cases. We therefore could not estimate the risk of brain cancer in more detail in relation to risk for specific brain cancer subtypes. Second, data on the accuracy of brain cancer diagnosis are not available in Taiwan and misclassification is possible. However, this misclassification is likely to be nondifferential (i.e., unlikely to be related parity) and therefore would tend to underestimate rather overestimate the true association. Third, Taiwan’s vital records and birth registration system covers only live births and exclude stillbirths and abortions. We were therefore unable to examine the possible role of gravidity on the brain cancer risk. Fourth, by design, our study focused solely on mortality among parous women. We were unable to examine the possible role of nulliparity on the risk of brain cancer. The generalizability of our findings is thus limited. Fifth, although the number of person-years are large, the number of of brain cancer deaths is relatively small (n=316). The power of this study was therefore relatively limited. The possibility that this is a chance finding also needs to be considered.