Gregory L. Katzman, Azar P. Dagher, Nicholas J. Patronas. Incidental Findings on Brain Magnetic Resonance Imaging From 1000 Asymptomatic Volunteers. JAMA. 1999;282(1):36–39. doi:10.1001/jama.282.1.36
Author Affiliations: Diagnostic Radiology Department, National Institutes of Health, Bethesda, Md.
Context Previous reports have discussed incidental disease found on brain magnetic
resonance imaging (MRI) scans that had been requested for an unrelated clinical
concern or symptom, resulting in a selection bias for disease. However, the
prevalence of unexpected abnormalities has not been studied in a healthy population.
Objective To evaluate the prevalence of incidental findings on brain MRI scans
obtained for a healthy, asymptomatic population without selection bias.
Design, Setting, and Participants Retrospective analysis of brain MRI scans obtained between May 17, 1996,
and July 25, 1997, from 1000 volunteers who participated as control subjects
for various research protocols at the National Institutes of Health. All participants
(age range, 3-83 years; 54.6% male) were determined to be healthy and asymptomatic
by physician examination and participant history.
Main Outcome Measure Prevalence of abnormalities on brain MRI by category of finding (no
referral necessary, routine referral, urgent referral [within 1 week of study],
and immediate referral [within 1 to several days of study]).
Results Eighty-two percent of the MRI results were normal. Of the 18% demonstrating
incidental abnormal findings, 15.1% required no referral; 1.8%, routine referral;
1.1%, urgent referral; and 0%, immediate referral. In subjects grouped for
urgent referral, 2 confirmed primary brain tumors (and a possible but unconfirmed
third) were found, demonstrating a prevalence of at least 0.2%.
Conclusion Asymptomatic subjects present with a variety of abnormalities, providing
valuable information on disease prevalence in a presumed healthy population.
A small percentage of these findings require urgent medical attention and/or
Unexpected abnormalities are occasionally discovered during brain magnetic
resonance imaging (MRI), usually in the setting of an investigation for some
other reason. The radiologist and referring physician are then placed in a
position of determining relevance of the abnormal finding and considering
its impact on the patient. To this end, decisions must be made concerning
the seriousness of the finding, including whether it is merely within the
realm of normal variation.
Studies have been described that attempt to report on the prevalence
of such incidental findings. This includes articles pertaining to discussions
of sinusitis,1- 3
white matter lesions,4,5 and even
pineal cysts.6,7 In each case,
data are offered addressing prevalence within certain populations, establishing
a foundation on which other MRI studies can be compared. All of these prior
reports establish statistical occurrence on MRI examinations that had been
performed for other reasons, yet not on healthy subjects.
Several investigational protocols throughout the various institutes
that make up the National Institutes of Health (NIH) use brain MRI analysis,
and many of these also require a healthy brain MRI database for comparison
purposes. Each protocol independently admits both patients and healthy volunteers,
and a participant history taking and a physical examination are performed
by a physician for both populations. Healthy volunteers are actively recruited
and are paid for their participation. Subjects with signs or symptoms are
excluded. Such MRI studies are widely used, and the topics of investigation
vary from measuring specific anatomic structures to complex-functional MRI
scans to whole-brain spectroscopy, all of which share a common denominator
in generating an initial diagnostic brain evaluation for a clinical review.
We recognized the wealth of information this situation affords in the
evaluation of true incidental findings in a population actively selected for
healthiness by both the history taking and physical examination. Thus, we
retrospectively analyzed brain MRI results for 1000 healthy volunteers and
attempted to categorize them based on the urgency of any findings. We discovered
that asymptomatic healthy subjects present with a variety of abnormalities,
and a small percentage of these lesions required urgent medical attention
and/or additional studies. We also compared a selected few of these abnormalities
with epidemiological data.
We retrospectively analyzed brain MRI scans for 1000 healthy volunteers
performed within a 14-month period from May 17, 1996, to July 25, 1997. Abnormalities
were categorized using the method chosen by Bryan et al8
for the Cardiovascular Health Study. This method classifies all findings into
4 basic categories: (1) no referral necessary; normal or findings common in
asymptomatic subjects (eg, sinusitis); (2) routine referral; findings not
requiring immediate or urgent medical evaluation, but should be reported to
the referring physician (eg, old infarction); (3) urgent referral required
within weeks of study for any abnormality that will need further yet nonemergent
evaluation (eg, low-grade astrocytoma); and (4) immediate referral required
(eg, acute subdural hematoma).
We evaluated brain MRI studies
that were often limited but contained at least T1-weighted and
T2-weighted imaging sequences. Proton density sequences were available
only occassionally and more advanced sequences were rarely available at the
time of interpretation. All studies were performed on General Electric Signa
scanners (Milwaukee, Wis). All protocols admitted subjects under an intramural
project approved by the institutional review board at the NIH (Bethesda, Md).
All studies were interpreted by board-certified radiologists who had a certificate
for added qualification in neuroradiology.
Of the 1000 patients included within this retrospective analysis, 546
(54.6%) were male and 454 (45.4%) female, with a mean (median) age of 30.6
(29.0) years (range, 3-83 years). Eighty-two percent of the studies were normal.
Abnormal scans accounted for 18% and were classified as follows:
151 (15.1%) without referral, 18 (1.8%) routine, 11 (1.1%) urgent, and 0 (0%)
immediate. Abnormalities within the no referral category included sinusitis
(n=132), age-related changes (n=12), solitary nonspecific T2 hyperintensity
(unidentified bright objects [UBOs]) (n=5), and mastoid/petrous fluid (n=4),
of which simple sinus disease constituted a vast majority. Abnormalities within
the routine referral category included old lacune (n=3), possible demyelinating
disease (n=3), choroid cyst (n=2), pineal cyst (n=2), tornwaldt cyst (n=2),
empty sella (n=1), nasopharyngeal cyst (n=1), hypothalamic lipoma (n=1), prominent
temporal horns (n=1), remote traumatic change (n=1), and scalp cystic lesion
(n=1). Urgent categories were arachnoid cyst (n=3), cavernous angioma (n=2),
benign but recommended additional imaging (n=2), low-grade oligodendroglioma
(n=1), pilocystic astrocytoma (n=1), low-grade glioma (unconfirmed) (n=1),
and aneurysm (unconfirmed) (n=1). There were no findings classified as immediate,
which was expected for a healthy volunteer population.
was noted in 13.2% of the participants. Overall, sinusitis was more prevalent
in the spring months of February, March, and April and maxillary sinuses were
more often affected.
Age-related changes occurred in 12 subjects
who were older than 55 years: 40% of the women and 60% of the men experienced
microvascular disease (mean [median] age, 73  years); 37% of the women
and 63% of the men experienced atrophy (mean [median] age, 74  years);
and 0% of the women and 100% of the men experienced old lacunar infarctions
(mean [median] age, 68  years).
Of great interest is the finding
within the urgent category of 2 confirmed and 1 unconfirmed primary central
nervous system (CNS) tumors. One of these lesions was subsequently resected
at the NIH, and pathologic analysis confirmed low-grade oligodendroglioma.
A second lesion was resected at a local institution and histologically graded
a pilocytic astrocytoma. A third lesion demonstrated MRI findings suggestive
of a low-grade glioma9; however, further evaluation
at the NIH was not performed, and follow-up care is not known to us.
As expected, a vast majority of healthy volunteers demonstrated a normal
MRI scan appearance. It is important to note the young age of our cohort because
this dramatically decreased the chance of revealing pathologic findings such
as metastases or those related to age, which are found in a much older age
group. In fact, there were only 12 patients available who had age-related
changes, and we believe it would be inappropriate to draw conclusions for
such a small sample size. The young age of the cohort most likely occurred
because experimental MRI studies are relatively demanding. The 2 best examples
are functional MRI scans and whole-brain spectroscopy; the former requires
dedicated intricate instructions as well as volunteer activity, and the latter
can take as long as 90 minutes within the scanner bore. Such activity can
be demanding and likely appeals to younger individuals more. Additionally,
the older population may not feel the need to participate in medical research
and/or may not need the small stipend paid to our volunteers. The ratio of
male to female volunteers was nearly even.
The NIH is a tertiary
care center with strict criteria for inclusion within either an experimental
patient protocol or a healthy volunteer study. Thus, any deviance from these
strict confines results in removal of the patient/subject. The referring tertiary
care center is notified of the abnormal findings for the patients; for healthy
volunteers their primary care provider is informed of the abnormality of concern.
Nevertheless, in both instances, patients/subjects do not return to the NIH
and thus are lost to follow-up. As the NIH generates nearly 1000 healthy volunteer
studies each year, it is untenable to pursue follow-up at the outside institution.
Our findings of paranasal sinusitis are significantly less than
the reported literature, which ranges from 41.6% to 49.2%.1- 3
We believe that our much lower percentage of 13.2% more accurately reflects
the true prevalence of sinusitis in healthy populations because our study
cohort was specifically chosen for healthiness, whereas prior studies have
interpreted sinusitis as incidental findings within a population already undergoing
brain MRI for some other reason. One can easily imagine that sinusitis could
account for signs or symptoms that might not initially seem localized to the
paranasal sinuses, for example, those masquerading simply as a headache as
an indication for brain evaluation. Additionally, the largest of these studies
had 325 patients, less than one third the number in our study. Our findings
that the maxillary sinuses are more commonly involved agree with results previously
have been previously described in a fairly large body of literature.8 Given the young cohort in our study, only 12 subjects
fell within this category of findings. A majority of these 12 patients had
been enrolled by the National Institute on Aging and likely had been actively
A large body of literature also exists concerning UBOs
within the white matter, which we refer to as solitary nonspecific T2 hyperintensity found in 5 subjects (0.5%). We caution against inferring
any conclusions from these data for 2 reasons. First, our young cohort puts
us at a disadvantage as the classic UBO is found in individuals older than
the mean age of our cohort. Second, the images available for interpretation
usually contained T1- and T2-weighted techniques, whereas
such solitary T2 abnormalities are usually best found on proton
density or fluid-attenuated inversion recovery techniques.
found 3 younger patients (0.3%) with many nonspecific focal T2
white matter hyperintensities. We classified these as possible demyelinating
disease (listing additional differential diagnoses within the report) and
all were lost to outside follow-up. Thus, we have no idea how many, if any,
of these subjects truly had a demyelinating disorder.
urgent referral category, several lesions of interest need to be discussed.
We found 3 patients (0.3%) with arachnoid cysts. Robinson10
states that arachnoid cysts are relatively uncommon and account for approximately
1% of intracranial masses. Unfortunately, the only literature discussion of
population prevalence is limited to findings from fetal/neonatal autopsies
that describe a 0.17% prevalence.11 We believe
our findings are within an acceptable range.
has been reported to occur at a 0.47% prevalence12;
however, in that report of 66 patients only 9 patients (0.06%) had true incidental
lesions, since all others had been referred for imaging investigation of neurologic
signs or symptoms. Thus, the acceptable range of prevalence is likely between
0.06% and 0.47%. Our finding of 0.2% falls within this range.
Intracerebral aneurysm prevalence has been described in the literature ranging
from 0.2% to 8.9%,13 with autopsy-based studies
yielding the lower and angiography-based studies yielding the higher rate
of prevalence. The mechanism of aneurysm formation is unknown and may be secondary
to the constant hemodynamic stress placed on arterial branch points over time.13 Thus, in our young cohort, our 0.1% rate (1 unconfirmed
aneurysm) is within an acceptable range.
Most interesting, and
potentially worrisome, is our findings in 3 subjects of suspicious primary
brain neoplasm. Indeed, 1 of these subjects was allowed to remain at the NIH,
and was admitted within an existing brain tumor protocol, and subsequent resection
resulted in histological verification of low-grade oligodendroglioma. Meticulous
preoperative physical examination of this volunteer by neurosurgeons failed
to uncover any signs or symptoms. A lesion in a second subject was resected
at a local institution with histological grading of pilocytic astrocytoma.
Finally, a third patient demonstrated findings suspicious for low-grade glioma;
however, follow-up evaluation is unknown. At a minimum, we can conclude that
our population of healthy volunteers included 2 subjects (0.2%) with an asymptomatic
and confirmed primary brain neoplasm. A previous report14
examining brain MRI scans from 3672 patients demonstrated no primary CNS neoplasms,
1 patient with known CNS lymphoma, and 19 meningiomas. However, all of these
individuals were older than 65 years and had been enrolled in a longitudinal,
population-based study of cardiovascular and cerebrovascular disease. It is
interesting to note that this study showed a 1.7% rate in the urgent category,
which is close to our rate of 1.1%.
Davis et al15
have described primary brain tumor incidences from the states of Connecticut,
Massachusetts, Missouri, and Utah derived from data that had been entered
into the Central Brain Tumor Registry of the United States. This group of
states comprises a population of approximately 16 million subjects with 8070
reported primary tumors, for an overall calculated incidence of 9.4/100,000
person-years (0.0094%). They report estimated completeness of reporting to
range from 90% to 99% for the 4 states. However, they also address the problem
of underascertainment, which they believe may be as high as 30% to 100%, and
note that this problem will ". . . require further work."15
These data were maintained by the American Association of Central Cancer Registries.
Preston-Martin16 has also described
the epidemiological characteristics of primary CNS neoplasms from Los Angeles
County, California.16 Her study evaluated reporting
and showed an overall calculated incidence of 10.4/100,000 person-years (0.0104%).16 This correlates well with the study by Davis et al.15 Additionally, this study broke down incidences within
age groups, with an overall incidence of 5/100,000 person-years (0.005%) for
ages 25 to 34 years, the range within which our mean ages both fall. Her study
also stated that ". . . there is often a significant underascertainment of
cases. . . ."16
of 0.2% for primary neoplasm is difficult to interpret. Given the following
definitions: prevalence=N with disease at time t divided by total N at time t. Incidence=n developing disease
divided by [n at risk×period of observation],
our data more closely approximate that of a prevalence and do not strictly
adhere to either definition. As described earlier, the reported incidence
for primary CNS neoplasm is 20- to 30-fold lower than our findings. We are
measuring closer to a disease prevalence, which may help explain the large
discrepancy with the reported literature. However, we also feel this is still
higher than what should be expected, and that the difference likely cannot
be explained by underascertainment alone.
Both articles discussed
earlier fail to discuss tumor prevalence in their populations. This raises
several questions: Is the rate of primary brain neoplasm actually much higher
than that reported in the literature? Do patients feign healthiness, perhaps
for a free evaluation of symptoms they have or for money? Are the admitting
history taking and physical examinations not rigorous enough? We are unsure
of the answer, and it may be a combination. We do warn that the true rate
of primary brain neoplasm may be higher than that reported, but heavily caution
that our findings are based on a much smaller sample population (1000) of
subjects compared with the millions evaluated in the aforementioned literature
and may merely represent a statistical anomaly. However, as the availability
of MRI becomes more uniform throughout all geographic regions, we feel the
number of abnormal findings relegated into an urgent referral pattern probably
will increase. This will likely also result in a corresponding increase in
primary CNS neoplasm diagnosis both at the clinical and histological levels.
We also caution that the literature discussions of anatomic lesion incidence
or prevalence are fraught with selection bias, because a vast majority of
imaging is performed for the evaluation of patient signs or symptoms. Thus,
it is not always straightforward whether an identified abnormality is the
causal agent or a true incidental finding, making it difficult to infer a
true population prevalence. In this article, we present data for 1000 subjects
who were actively selected for their healthiness, which should free our findings
from such bias.