Ethical and practical considerations in the management of incidental findings in pediatric MRI studies.

The observation that the onset of schizophrenia and other serious neuropsychiatric disorders typically occurs during late adolescence has driven considerable research toward understanding maturational changes in brain anatomy during this time period in healthy adolescents. Magnetic resonance imaging (MRI) research is considered safe and noninvasive and thus appropriate for use in children and adolescents. MRI protocols that have included children and adolescents have been typically viewed as entailing "minimal risk" by institutional review boards. "Minimal risk" as defined through use of the Common Rule means "that the probability and magnitude of harm or discomfort anticipated in the research are not greater in and of themselves than those ordinarily encountered in daily life or during the performance of routine physical or psychological examinations or tests."

The main risks from participation in these studies are claustrophobia, acoustic noise, and the identification of a previously undiagnosed, untreatable serious medical disorder. Other highly unlikely risks include tissue heating from radiofrequency energy deposition (less than 1°), static magnetic field risks (e.g., with metal implants), and paresthesias from gradient switching. The risk of claustrophobia can be minimized with preparatory training and education to reduce patient anxiety and also by scanning children when they may be naturally sleepy. To minimize acoustic noise, subjects are offered earplugs or audiovisual entertainment while being scanned. Last, despite the small prevalence of undetected neurological problems in healthy pediatric volunteers, some authors have suggested that the presence of any significant clinical findings is a matter of medical and bioethical concern. Consequently, they have advocated for the routine involvement of a neuroradiologist in research MRI studies, particularly those involving children, to ensure the proper detection and determination of the significance of MRI abnormalities.

This approach can be problematic because clinical neuroradiologists typically read scans of children who are receiving diagnostic workups to rule out some type of brain abnormality, and they are trained to identify pathology. Because of this training and concerns regarding medical malpractice, there is a tendency on the part of neuroradiologists to err on the side of safety and to report all observed findings.

Recently, there have been preliminary reports regarding the observation of unexpected MRI structural anomalies that may or may not have functional significance in healthy children and adolescents who were research participants. On visual inspection, these MRI scans look abnormal to the trained reader, but without an associated functional consequence, the interpretation and appropriate management of these findings is problematic for the principal investigator. In part, the difficulty in the interpretation of these scans reflects the dearth of normative data in healthy children and adolescents with regard to the spectrum and prevalence of brain structural deviations. Initial reports of incidental MRI findings in healthy children and adults have raised important questions regarding the need for consensus in terms of who should read the MRI scans of research participants and applicable safeguards to ensure that appropriate referrals are being made if an abnormality is identified.

To our knowledge, there have been relatively few published data regarding the prevalence, significance, and practical management of unexpected MRI findings that are detected by a clinical neuroradiologist in healthy children and adolescents participating in research MRI protocols. These data are needed to properly characterize research risks from both legal and ethical standpoints for child and adolescent psychiatrists conducting MRI research. These considerations provided an impetus for the present study. For the past 3 years, our group has conducted diffusion tensor imaging studies in adolescents with early-onset schizophrenia to look for evidence of a white matter component in the pathophysiology of the disorder. In this article, we retrospectively review the MRI reports from 60 children who participated in these studies as normal controls who were presumed to be "neurologically healthy" by clinical history.

DISCUSSION

In this case series, we found that 8 (13%) of 60 healthy pediatric volunteers had unexpected MRI abnormalities that were detected by a clinical neuroradiologist in children who were participating as healthy volunteers in a research MRI study. These data are comparable to previously published data. Although the majority of these unexpected findings required no additional follow-up other than communication of the structural anomaly to the primary care physician, three (5%) subjects were found to have a potentially clinically significant neuroradiological problem that required additional clinical follow-up. This number may be higher than what would be encountered in most research settings because of the inclusion of T2-weighted and FLAIR sequences in our research protocol.

Research protocols at many institutions do not acquire these sequences because the sequences of greatest research value for functional MRI (fMRI) and structural MRI (sMRI) are usually T1-weighted and echo-planar sequences. Thus, the abnormalities reported on the T2-weighted sequences we acquired for our diffusion tensor imaging protocol likely would not have been detected at many centers focused on either fMRI or sMRI because they would not have been visible on the T1-weighted sequence and the quality of echo-planar images is not adequate for clinical readings. This would have reduced the rate of apparent clinical abnormalities to 0%. However, we chose to include a T2-weighted sequence because, for diffusion tensor imaging protocols, this sequence is necessary for distortion correction and segmentation purposes. Also, we decided to include an additional 4-minute FLAIR sequence that enhanced the potential diagnostic value of our research protocol to provide some clinical benefit primarily to the adolescents with schizophrenia populations that we were scanning, many of whom were in the first episode of their illness.

In these three cases, a follow-up MRI and neurological evaluations clearly ruled out the possibility of a tumor, a vascular malformation, and a demyelinating disease. In this study, the specificity of detecting clinically significant abnormalities based on the initial clinical reading and the ensuing diagnostic workup was poor (i.e., zero), despite the inclusion of pulse sequences (i.e., T2-weighted, FLAIR) that were of much better quality for clinical readings than those that are acquired in most research settings that focus on fMRI and sMRI. One healthy volunteer was clearly distressed after the need to undergo a follow-up MRI to rule out the possibility of a vascular malformation was disclosed. As a result, this volunteer chose not to participate further in our research study.

In another case, although isolated hyperintense lesions on T2-weighted images have been associated with neurofibromatosis and psychiatric disorders, the significance of the anomaly for this subject was unclear, prompting the recommendation by our neurologist to obtain a follow-up MRI scan to examine progression. At the 2-year follow-up, the initial abnormality appeared stable, and thus it was deemed to be of no clinical significance. In this case, the provision of a diagnosis of a structural anomaly without some type of immediate intervention or treatment resulting in resolution of the problem was problematic because the disclosure generated anxiety in both the child and the parent. Overall, these clinical vignettes raise questions regarding how best to communicate to subjects and their parents unexpected MRI findings in an effort to minimize anxiety while ensuring that recommendations regarding clinical follow-up are implemented.

The most obvious implication of these data is that a proactive approach seems most prudent in which the possibility of incidental findings should be explained to research subjects as part of the informed consent process. Beginning this conversation before the MRI scan will help to facilitate the communication of an incidental MRI finding to a parent/child after the MRI procedure. As part of this discussion, the subject and the parent could be advised that the images acquired for the research protocol generally would not be adequate for a clinical reading and therefore may falsely suggest the presence of an abnormality and/or may not allow for optimal detection of actual abnormalities. Also, some of the detected abnormalities may turn out to be benign, or normal variants and may reflect variation because of ethnicity, nutritional, genetic, or some other type of environmental factor; they are not a cause for concern and knowing that these features are present in their MRI for each subject may help prevent problems when they are seen on later MRI. However, on occasion, a neuroradiologist may observe abnormalities that may be within normal limits, or outside normal limits, and that may require clinical follow-up. It also could be explained that from a time standpoint, it is not feasible to obtain all possible brain images during one single scanning session that are optimized to finding all different types of brain abnormalities.

As part of the informed consent process, it may also be conceivable that subjects would elect an option that would allow the parent/subject to opt out of being informed of incidental findings altogether unless they were considered to be of a serious nature requiring additional clinical follow-up. The scenario in which a child's wishes may differ from those of the parents is a common question in many research settings involving consent and notification. However, in this example, it seems that both the guardian and the child participant may have to provide informed consent/assent to this option to make research participation feasible. As with any research procedure, both the guardian and the child participant would have the option to change their mind in the future, and the signature on the consent form would not absolve the researcher of all liability because no consent form could be written that would remove the possibility of negligence on the part of the principal investigator.

At a recent National Institutes of Health meeting, it was concluded that there should not be a burden that every research imaging protocol include a diagnostic set of images. Because the scans typically available in most research imaging studies do not comprise a complete set and are not therefore diagnostic, there was less consensus on the issue as to whether every MRI scan should be clinically reviewed by a neuroradiologist, or whether a neuroradiologist should only become involved in cases that were noted to have obvious abnormalities by a member of the research team. An ethical question raised by the data presented herein is whether the benefits provided by having a neuroradiologist read all of the scans in terms of ruling out the presence of any significant clinical findings were offset by the anxiety created by disclosing to some subjects that they had a structural anomaly that required a follow-up MRI examination that then later turned out to be a false positive. In addition, recommendations for clinical follow-up should not be considered benign altogether because each additional procedure also entails specific risks (e.g., gadolinium injection, economic costs).

As part of their deliberations, institutional review boards should consider the following issues: On the one hand, research studies that do not have scans clinically read by a neuroradiologist run the risk that unanticipated findings in subjects may go unrecognized and thereby leave subjects without appropriate referral. From a scientific perspective, the inclusion of a diagnostic set of images and the involvement of a neuroradiologist in research studies examining structural brain abnormalities is advantageous in terms of the interpretability and validity of findings because all subjects with intracranial abnormalities would be excluded. However, as previously emphasized, the probability of detecting a clinically silent lesion from research protocols that include T1-weighted images alone would be highly unlikely particularly if the clinical evaluation before the scan was normal.

On the other hand, as shown in this case series, the involvement of a neuroradiologist in MRI research studies entails an increased likelihood of making children and their parents frightened by an excessive number of recommendations for follow-up medical evaluation or imaging procedures. For example, motion artifact or signal voids from susceptibility effects at tissue interfaces on T1-weighted images often are reported by radiologists as a "possible lesion of uncertain clinical significance," and radiologists will typically recommend further diagnostic evaluation in these situations to protect themselves from medicolegal risks.

Although we did not formally test the expectations of the healthy volunteers who participated in this study, we assume that the families/participants volunteered in part for the compensation provided, the opportunity to do something altruistic, and with the expectation of provision of some assurance of a "clean bill of health" provided by the principal investigator to the research participant. Because of the therapeutic misconception, it is difficult for the lay public to understand that most MRI researchers are primarily focused on testing specific hypotheses and looking for specific things and that they are not trolling for potential clinical problems in the service of helping the research participant. To discourage subjects and their parents from study participation who are primarily interested in obtaining some type of clinical benefit, Illes and colleagues (2002) have suggested including language in the consent form to notify subjects that the images will not be made available to the family for diagnostic purposes because they do not represent a proper clinical MRI series. However, merely adding such language into the consent form is insufficient and time must be taken with the family to explain the absolute lack of clinical meaning of the procedure as part of the informed consent process with the research participant and the parent or guardian.

In terms of institutional policies regarding the management of incidental findings, the International Conference on Harmonization Relating to Good Clinical Practice (ICH topic 6), section 4.3.2 under investigator responsibilities states: "During and following a subject's participation in a trial, the investigator/institution should ensure that adequate medical care is provided to a subject for any adverse events, including clinically significant laboratory values, related to the trial". At our institution we apply good clinical practice guidelines broadly and use these guidelines as the basis for our research compliance program.

Limitations

There are a number of limitations that should be emphasized for this study. Healthy volunteers were recruited for this study in response to advertisements posted in pediatricians' offices or community centers rather than using a more sophisticated epidemiologically based sampling methodology. Although the mean IQ for the normal controls from our sample was 107, which is within the average range, all of the children and adolescents were from higher parental socioeconomic status backgrounds. We would hypothesize that parents of children who are more educated would be more willing to allow their child to participate in an MRI research study. Therefore, the selection or ascertainment biases involved in this study may limit the generalizability of these data. In addition, the sample, although large for most pediatric MRI studies, is still relatively small to detect rare events; thus, these results will need to be replicated in larger studies of normative brain development that are ongoing to better evaluate the advantages and disadvantages of having a neuroradiologist routinely read all research MRI scans. With a larger sample, it would also be possible for a health economist to determine the incremental cost-effectiveness of two different approaches surrounding the detection of incidental findings: (1) a member of the research team having responsibility for examination of the MRI scans for incidental findings and radiologist involvement only if consultation deemed necessary versus (2) a board-certified radiologist providing a clinical evaluation of all research scans.

Clinical Implications

In this article, we have discussed whether having MRI scans read by a clinical neuroradiologist serves to better protect research subjects or inadvertently to harm them by informing them that they may be at risk of a medical disorder when they actually have none. From a legal perspective, by telling potential subjects that we may see "something" on their MRI scans, we are in essence telling them that we are looking for "something" and both the neuroradiologist and the research investigator assume some potential "clinical responsibility." As suggested by a recent NIH workshop group, the neuroradiologist should provide information to the principal investigator about incidental findings and should inform the principal investigator about what information should be conveyed to the participant's primary medical care provided. The clinical neuroradiologist and the research investigator must provide appropriate diagnostic and follow-up care, respectively, that conforms to the standard of care within their communities. There is nothing that could be written into the consent form that would take away the responsibility of either physician to provide this basic standard of care for research subjects. However, to limit liability the principal investigator could propose transferring the responsibility for the determination of the significance of any discovered incidental findings and the need for follow-up tests to the participant's primary medical care provider. In this scenario, the investigator would need the subject's explicit, advance permission. If a primary care provider does not exist, then appropriate care would need to be arranged by the principal investigator.

For clinicians who are referring subjects for participation in research studies, it should not be assumed that all MRI scans are of similar value for clinical/diagnostic purposes. For fMRI and sMRI protocols, clinical readings of the most commonly acquired research scans, T1-weighted (and echo-planar) images, at best can only provide information about large, space-occupying lesions and the presence or absence of hydrocephalus. The presence of these abnormalities would almost always be detected by a careful review of symptoms and neurological examination. In contrast, for diagnostic purposes, a routine MRI protocol would include T1-weighted, a double-echo T2-weighted and FLAIR sequences. These sequences were acquired as part of our DTI research protocol.

Although not the focus of this report, in our clinical sample, the clinical neuroradiologist noted abnormalities in 6 (8%) of 77 patients that required further clinical evaluation. Only one abnormality, which consisted of a subcentimeter lesion in the left lateral ventricle, required clinical follow-up. This patient chose to pursue these additional clinical examinations elsewhere and thus no follow-up data are available. The remaining nonspecific findings in the other five patients included (1) a small punctate focus of increased FLAIR signal in the left parietal subcortical white matter, (2) prominent pituitary gland, (3) mild to moderate small punctuate foci of increased FLAIR signal intensity noted in the anterior bilateral frontal subcortical white matter, (4) nonspecific foci of T2 prolongation identified in the left frontal region, and (5) an oval soft-tissue focus of increased T2-weighted signal intensity present in the region of the right temporalis muscle/lateral frontal scalp.

Last, there have been growing concerns raised regarding direct-to-consumer marketing of self-referred imaging services in both print advertisements and informational brochures. These types of solicitations are problematic because some of the claims made in these ads, for example, that psychiatric disorders may be diagnosed with imaging tools, are not supported by fact. Also, the comprehensive balanced information that is vital to informed autonomous decision making typically is not provided in these ads to parents of children with severe emotional disturbances and/or learning disabilities who may be looking for answers to their child's problems. Adult patients with chronic schizophrenia have typically been found to have an increased prevalence of incidental brain abnormalities. Clinicians faced with parents requesting consultation regarding the significance of these findings must appropriately advise parents that the correct interpretation of many of these findings (e.g., generalized loss of brain parenchyma, enlargement of lateral ventricles) is difficult at present because of the insufficient amount of available pediatric normative data.