The results indicated that there were differences in trend patterns between the two regions. The Serrana Region had higher mortality rates and an increasing trend in mortality over the period analyzed (1996–2010). In contrast, an opposite trend occurred in the Metropolitan area. Mortality rates were lower and declined during the study period, although the decrease was not statistically significant.
Monteiro and Koifman  reported an increase in brain cancer mortality rates in Rio de Janeiro between 1980 and 1998 in individuals >65 years of age. Legler et al.  analyzed brain cancer mortality rates in the United States between 1975 and 1999, and reported a stable distribution of mortality rates, except in the age group between 64 and 74 years of age. This group had an increase of 5.5% in the EAPC between 1979 and 1995 . In the Umbria Region of Italy, Stracci et al.  reported an increasing trend in brain cancer mortality rates of 2.33% (95% CI: 1.42–3.23) in males and 1.78% (95% CI: 0.62–2.95) in females.
The increases in brain cancer incidence and mortality rates that have occurred in recent decades may be attributed to improved diagnostic capability that has resulted from the use of computed tomography (CT) and magnetic resonance imaging (MRI). Population aging has likely also contributed to this change, because age represents an important risk factor for intracranial tumors [32–34]. However, new technologies and aging do not fully explain the increases in incidence and mortality, and there may also be a significant contribution from environmental risk factors .
Differences in the magnitude of brain cancer mortality rates observed in this study cannot be explained by greater access to MRI and CT scans. The magnitude of the adjusted mortality rates in the Serrana Region is somewhat higher than the rates in Rio de Janeiro, which has greater access to these diagnostic tools. One hypothesis for the dissimilarity is differences in patterns of exposure to distinct environmental carcinogens between the two regions.
The result of this study indicated that there was a statistically significant age effect on the distribution of brain cancer mortality rates in both regions. Age is an important risk factor in the development of several types of tumors. The number of cell divisions increases during human aging. During cell division errors in DNA replication occur that are critical for the formation of mutations. When these mutations occur in DNA repair mechanisms, they can result in the development of tumors . Flaws in DNA replication can also be induced by specific environmental agents .
The most recent birth cohorts in the Serrana Region had higher RRs. This effect may reflect changes in exposures to environmental agents that occurred after 1950, and that have been present since then. Environmental factors likely contribute to the risk of developing brain cancer. Many substances are inducers or promoters of carcinogenesis, including several pesticides [38–41].
The hypothesis for this difference in RR among the birth cohorts of the two regions accounts for differences in patterns of environmental exposures. The greater RRs in the 1980s cohorts may reflect exposures that occurred in childhood, because those individuals were ≤30 years old when they died. Exposure to pesticides in utero and during childhood is a potential risk factor for the development of brain cancer [42, 43]. Humans may be exposed to pesticides from several sources, including pesticides present in food and in agricultural and residential areas . The timing of the exposure during development is also important, because specific developmental periods during childhood are more sensitive to the biological effects associated with pesticide exposure . Exposure during these periods may significantly contribute to the risk of development of cancer in adult life, but the causal relationships are not clear.
Compared with the Metropolitan Region, younger patients in rural regions may not have the same level of access to early and accurate diagnosis and effective treatment. Survival rates of rural patients may be lower because of delayed diagnosis and delayed transfer to the more developed cancer hospitals in the cities. Additionally, the results in Table 1 indicated that age is the strongest risk factor. Individuals <35 years in the Serrana Region and <25 years in the Metropolitan Region had the lowest mortality rates. In the Serrana Region, individuals from the most recent birth cohort had four times greater mortality rates, compared with those born in 1945–1949 (referent birth cohort).
Over the past 30 years, the Serrana Region has gone through a process of agricultural modernization . This region is the main agricultural area in Rio de Janeiro, produces mainly vegetables, fruits and flowers, and employs the greatest numbers of workers engaged in agricultural activities in the state [12, 22]. According to Brazilian Institute of Geography and Statistics data, large amounts of pesticides are used to grow fruits, vegetables, and flowers. The 1996 volume of pesticides sold in the Serrana Region represented approximately 50% of the total sales volume in the entire state .
Consumption of pesticides in Brazil increased from 600 million liters to 850 million liters between 2002 and 2011. The number of commercialized chemicals increased from 468 in 1995 to 600 in 2003. Per hectare consumption of pesticides increased from 3.2 kg to 3.6 kg between 2000 and 2009. In the Serrana Region, pesticide use has been high since 1986, which suggests that the population has been exposed to high levels of these chemicals over the last three decades. Considering the latency period between exposure and cancer diagnosis, it is reasonable to propose that the high use of pesticides in this region could have contributed to increases in the occurrence of diseases related to pesticides, including brain cancer [47–49].
Most of the pesticides used in horticulture, and fruit and flower cultivation are members of the organophosphate and carbamate classes of pesticides. Over the last few years, the carcinogenesis mechanisms associated with chemical induction and promotion of tumors by chemicals has been well-studied. Organophosphate and carbamate pesticides have two possible mechanisms of carcinogenesis. One mechanism is based on genotoxicity (ability to react with DNA) and the other is based on epigenetic mechanisms (changes that alter genetic expression without modifying the DNA sequence) . In vitro evidence indicates that organophosphate pesticides induce DNA mutations and methylation. The herbicide paraquat promotes changes in histone acetylation in cell culture [51–53].
Brain cancer in the Serrana Region should be more investigated further. Other studies have found that farmer and resident rural populations have high estimates of risk of death from specific cancers, especially brain cancers [54, 55]. Exposure to pesticides may have an important role for the development of brain cancer, as indicated by the mortality rates that were found in our study.
Our results should be interpreted cautiously because ecological studies can be affected by inherent design limitations . A common limitation of studies that use death certificate data is the accuracy of the mortality statistics. However, in a Rio de Janeiro-based study, Monteiro et al.  reported an accuracy of 90.1% in the reporting of death from brain cancer. In the Serrana Region, data on deaths from brain cancer had a positive predictive value of 90% . The ratio of the reported deaths in Chapter 18 (Sign Symptoms and Abnormal Findings in Physical Examination and Laboratorial Works) was 4.95% during the study period, and values <6% indicate good record quality . Another study limitation is inherent to uncertainties attributed to the nonidentifiability of the models [15, 28]. The three components age, period, and cohort are linear, and it is impossible to simultaneously estimate all three effects in the regression models. We used a method proposed by Holfrold to account for this problem .
This original study detected differences in the epidemiological patterns of brain cancer. Internationally accepted variables were used to study the distribution of disease (e.g., the distribution of mortality by age group (age effect), calendar year of death (period effect), and birth year of the deceased (cohort effect). This approach enabled us to generate hypotheses about the contribution of different environmental factors that may explain regional disparities in the distribution of mortality from brain cancer.
This study contributes to the understanding of ecological risk factors for death from brain cancer. The age-period-cohort model proved to be an efficient analytical method and found important differences in mortality patterns that suggest that there were differences in exposure between the two regions. We also found that there was a significant cohort effect, which suggested that residing in an agricultural area during early life increased the risk of mortality. This result supports the hypothesis that environmental exposures are determinants in mortality from brain cancer. Other studies of this population should be prioritized to determine the individual factors that are associated with the development of cancer.