Population-based epidemiologic studies

Population-based studies are generally considered more accurate and less biased than the more limited clinical or autopsy-based series. Few population-based studies focusing on metastatic tumors of the brain have been reported in the literature, and most of these studies are decades old. Guomundsson published the results of a population-based study performed in Iceland reviewing the incidence of central nervous system (CNS) tumors from 1954 to 1963. The annual incidence of metastatic and primary brain tumors was reported to be 2.8 and 7.8 persons per 100,000 population, respectively. The incidence proportion (defined as the number of patients with metastatic brain tumors by the number of patients with primary brain tumors in that particular population) of metastatic brain tumors in all patients with primary systemic malignancies in that study was less than 20%.

Percy and colleagues reviewed the data from 1935 to 1968 in Rochester, Minnesota, and found a much higher incidence of 11.1 per 100,000 population; however, the study group was mixed, with some patients being diagnosed clinically and others at the time of autopsy (70% of patients) and some with tumors confirmed pathologically and others diagnosed purely based on imaging. A study from Finland evaluating patients with metastatic brain disease from 1975 to 1982 reported the annual incidence of brain metastases and primary brain tumor at 3.4 and 12.3 persons per 100,000 population, respectively; brain metastases comprised 18% of all CNS neoplasms.

In an American survey of intracranial neoplasms, Walker and colleagues used hospital discharge records from 157 hospitals across the United States from 1973 to 1974 and estimated the annual incidence of metastatic and primary brain tumors to be 8.3 and 8.2 persons per 100,000 population, respectively; the incidence proportion of brain metastases was 51%. However, only 20% of all the cases reviewed in this study were pathologically verified. Counsell and colleagues identified 122 neuroepithelial primary brain tumors and 214 metastatic brain tumors in their population-based study of the Lothian region of Scotland from 1989 to 1990. They reported a yearly incidence of metastatic brain tumors of 14.3 persons per 100,000 population.

Materljan and colleagues reviewed hospital records in Labin, Croatia, from 1974 to 2001 and found a yearly incidence of metastatic and primary brain tumors to be 9.9 and 11.8 persons per 100,000 population, respectively. Barnholtz-Sloan and colleagues performed a population-based review of the Metropolitan Detroit Cancer Surveillance System from 1973 to 2001 (metro population approximately 4.5 million) and found the incidence proportion of brain metastasis among all patients with systemic malignancies to be 9.6%; however, this study was limited to the major types of cancer (lung, melanoma, breast, renal, colorectal). Smedby and colleagues used the Swedish national population-based health care registers from 1987 to 2006 and found that the annual incidence rate of hospitalization for brain metastases doubled during this period from 7 (in 1987) to 14 (in 2006) persons per 100,000 population. Schouten and colleagues used the Maastricht (Netherlands) Cancer Registry from 1986 to 1995, which covered 95% of the patients in this region, and identified 2724 patients with primary cancer. These patients were followed up until 1998, and 8.5% of them developed subsequent brain metastases, with 72% of the brain metastases occurring within the first year after initial primary cancer diagnosis.

Although population-based epidemiologic studies are considered better than clinical and autopsy-based studies, there are still some notable limitations and biases that are inherent to these types of studies. These defined populations and hospitals being studied are subject to (and limited by) the regional referral patterns, regional access to health care and cancer treatment, and the inherent sampling biases of the pathology of that region. In addition, no 2 population regions have equivalent treatment expertise. Slight variations in clinical aggressiveness in obtaining diagnostic imaging and/or biopsies or even the frequency of autopsies potentially affect the reported incidences of metastatic tumors.

Hospital records, registries, autopsies, and diagnostic imaging studies are subject to the inherent sampling biases, referral patterns, and treatment preferences of that particular region. Studies based on the above-mentioned data sources often underdiagnose the incidence of brain metastases. Many metastatic brain tumors are asymptomatic and are never diagnosed. In addition, patients with end-stage systemic cancer who develop neurologic symptoms near the end of life but are not clinically suitable for treatment given their systemic disease burden are never diagnosed. Up to one-third of all brain metastases are diagnosed at autopsy ; however, autopsy rates have significantly declined over the last 3 decades. Furthermore, there is variability in the definition of a metastatic brain tumor in that some studies consider all brain masses in a patient with a history of cancer to be a metastatic brain tumor, whereas other studies include only pathologically verified masses ( Table 1 ).

Cancer registries are frequently used in population-based studies to calculate incidence rates. Cancer registries have many advantages over other databases. They are usually confined to a particular state, region, or other defined location; follow all patients with a cancer diagnosis prospectively; record information with regard to status and treatment of cancer and death; and can calculate yearly incidences of cancer diagnoses for the particular population they cover. Smaller population- and hospital-based registries can interact and cooperate with each other and larger registries to cover a larger population size. Cancer registries also have limitations. They do not account for patients living within a defined region who seek medical care outside of the boundaries of the registry. In some American states, the population can be dynamic, which can affect the calculation of an accurate incidence. Further, cancer registries typically focus on and record only the primary cancer histology and primary site of the cancer and frequently do not contain information about metastatic brain tumors. These registries have also been shown to contain the International Classification of Diseases Ninth Revision coding errors and imprecise diagnostic codes. For example, a patient presenting with brain metastasis from a primary breast cancer is usually classified as having recurrent breast cancer rather than a separate brain metastasis. For these reasons, population-based and clinical studies can never be exact, and they typically underestimate the true incidence of metastatic brain tumors.

Clinical studies

Clinical studies have shown a high degree of variability in their reported incidences of brain metastases and vary significantly in the size and definition of the population being studied. Kawahata and Ohtomo reported on a cohort of elderly patients (median age, 77.5 years; range, 65–88 years) hospitalized with brain tumors in Japan from 1973 to 1987. They identified 322 pathologically confirmed brain tumors with an overall frequency of brain metastases of 5.8%. A study by Grant and colleagues reviewed imaging files, hospital records, and cancer registries in the Lothian region of Scotland from 1989 to 1990, which includes a referral population of 1.2 million patients. They found an annual incidence of intracerebral tumors (both primary and secondary) to be 21.4 per 100,000 population. In their study, 57% and 43% of the patients had metastatic and primary brain tumors, respectively.

Surgical series are an incomplete data source for brain metastasis information because they depend on patients referred for surgery and do not account for the metastases that are not treated surgically. Radiographic series are also incomplete sources to study the epidemiology of brain metastases because routine screening of the brain in patients with asymptomatic cancer is not typically performed except in certain lung cancer types (eg, small cell lung cancer) and they are also subject to the selection bias of referral patterns. Clinical studies based on hospital records are limited by selection bias and often use discharge diagnoses, which can be incorrect or nonspecific. For example, Walker and colleagues found that approximately 10% of these discharge diagnoses lacked specificity using terminology such as probable brain tumor, brain tumor, rule out brain tumor, and suspected brain tumor.