|Home | About | Journals | Submit | Contact Us | Français|
Background: Incidentalomas are findings on an imaging test done for other reasons, for which there are no matching symptoms in the patient. They are common in the adrenal gland, pancreas, liver, and thyroid, among other sites. Incidentalomas are a problem because we have a limited understanding of their natural history: it is difficult to know how much of a threat they pose to individual patients. An observational registry that would allow a systematic study of thyroid incidentalomas could reveal their natural history and the effect of detection on patients' lives, as well as document the cost to the healthcare system. A registry would help to determine which incidentalomas could be monitored and which require action.
Study design: A cohort study was conducted, with case identification via radiology imaging reports with follow-up through a minimum of one year post-identification.
Results: In one year, >109,000 imaging studies were performed that might reveal an incidental thyroid finding (computed tomography scans of the neck or chest, magnetic resonance imaging of the neck, plain x-ray of the chest, non-thyroid directed ultrasound of the neck, positron emission tomography scan, or myocardial perfusion scan). A total of 125 patients were identified as having a thyroid nodule, with a <1% reporting rate among eligible imaging studies, much lower than other published estimates of incidental thyroid nodule prevalence on imaging. Of the 125 nodules, 46 had been previously identified (were not “new”). Of the 79 patients with new nodules, more than half (44; 53%) were not notified of the finding. The approved study design allowed contact only with those who had been clearly notified of their thyroid nodule. Among those who could be reached, many did not recall the finding (6/15; 40%). Of those who did recall the finding, none self-identified it as an incidentaloma.
Conclusions: There are serious logistical and ethical hurdles to developing observational registries of incidentalomas, as well as threats to data validity because incidentalomas are incompletely identified, reported, and acted upon. Solutions commonly used to optimize data quality for registries would increase reporting, but could potentially overwhelm the healthcare system and harm patients. A novel interventional design that is proposed here for future work may facilitate both study and amelioration of the problem.
Incidentalomas are radiologically detected findings noted on an imaging test done for other reasons, and for which there are no matching symptoms in the patient. The term “incidentaloma” was first coined in 1982, when computed tomography (CT) scans became available (1). Adrenal gland findings, previously not typically discovered until very large, were suddenly being identified on the CT scans of people without symptoms related to their adrenal gland, creating management dilemmas for those called upon to investigate. Incidentalomas are now recognized to be common in many solid organs, and systematic management has been recommended (2). Incidentalomas are a particular problem in the thyroid gland because nodules are so common. By 50 years of age, half of all people have at least one thyroid nodule. By 90 years of age, virtually everyone has a nodule (3). In recent years, the detection of these nodules and of subsequent cancers, often through incidental detection on imaging done for other reasons (4–7), has led to the skyrocketing of incidence rates of thyroid cancer, without a significant change in mortality (8).
The incidentaloma problem as a general issue is a growing challenge for health systems because we have a limited understanding of the natural history of incidentalomas, which makes it hard to determine how much of a threat they pose to individual patients. Furthermore, the commonly held belief that early detection leads to better outcomes tends to promote intervention (9). However, many do not understand that early detection leads to longer estimates of survival because there is earlier knowledge of the finding, but does not necessarily decrease mortality—the number of people ultimately dying because of the pathologic process (10). Because of these problems, the solution remains unclear. The potential burden on patients is broad, and includes investigations, treatments, costs, and psychological effects. Ideally, research should be aimed at understanding and mitigating this burden. One approach to this would be to apply the registry model that has been successfully used in other diseases. Yet, the feasibility of a registry approach for thyroid incidentalomas remains unknown.
This study sought to develop an observational registry that would allow systematic study of radiologically detected incidentalomas of the thyroid. The goals were to understand their natural history, study the effect of incidentaloma detection on patients' lives, and document the cost to the healthcare system. The ideal registry envisioned at the outset of the project would capture all incidentalomas, permit access to the medical record for determination of downstream testing brought about by the incidentaloma, and allow direct contact with all affected patients over time. It was hoped that a general methodology for these types of registries would be developed, which might be expanded to cover incidentalomas in other organs in later iterations. The project met with many challenges: ethical, logistic, and design related. This experience is reported, and an alternative approach is suggested that may help advance study and amelioration of this problem.
The study was performed at Dartmouth Hitchcock Medical Center, Lebanon, New Hampshire, a rural tertiary care facility. It is a level I trauma center, and a National Cancer Institute–designated Comprehensive Cancer Center. The Department of Radiology, from which all cases included in the study were drawn, maintains its own electronic database of radiology reports created for every imaging test performed at the medical center. The database is searchable by individual words using SQL programming language. An algorithm for case identification was iteratively developed. By executing a SQL language command covering one month of eligible studies, and then checking a 10% random sample of excluded studies to see if any thyroid findings in the body or impression section were missed, the best search strategy was formulated. The most effective algorithm allowed for the use of the words “thyroid nodule,” “thyroid mass,” or “thyroid cyst” in the “body” or “impression” sections of the radiology report, but not in the “indications for imaging study” section of the report. Using this algorithm, a case was sometimes included in error (e.g., the imaging study was intended to study the thyroid and so not eligible for inclusion), but no cases were missed.
This was to be an observational cohort of persons identified as having an incidentally detected thyroid nodule based on the radiology report from eligible imaging studies. The initial study design called for identification of every patient with an incidentaloma from the radiology database (“case identification”), followed by personal contact with each one of those patients via phone survey, to determine their level of awareness about their incidental finding (“registry eligibility determination”). Patients who were found to be aware of their incidentaloma during that phone contact would be included in the portion of the registry designed to capture patient-reported outcomes. Patients who were unaware of their finding would not be informed of it via the phone contact, but would be followed only through chart review for evidence of any downstream events related to the incidentaloma.
The Human Subjects Research Institutional Review Board (IRB) at the institution would not permit every patient with an identified incidentaloma to be contacted, citing concerns that patients who had not been told of their finding might become aware of their finding through the process of determining by phone their level of awareness of their full radiology report. They would only allow people to be contacted who had clearly been notified of their finding. Though concerned about bias from leaving out patients who might know about their finding, even though it was not documented in the chart, an alternate design was developed in which only patients who had clearly been notified of their finding based on evidence in the medical record would be personally contacted and included in the registry.
To be identified as a case, a patient had to have had a CT scan of the neck or chest, magnetic resonance imaging (MRI) of the neck, plain x-ray of the chest, non-thyroid directed ultrasound of the neck, a positron emission tomography (PET) scan, or a myocardial perfusion scan between July 1, 2011, and June 30, 2012. These imaging tests were chosen because it had been established from prior research that incidentally detected thyroid lesions could be found on all of these imaging modalities (4). Patients had to be English speaking, and had to be older than 18 years of age. To be considered as an incidental thyroid finding, the body or impression section of the imaging report had to contain the words “thyroid mass,” “thyroid nodule,” or “thyroid cyst,” and the indication for the test could not contain the word “thyroid.”
After completing case identification, next thyroid findings that were not new were removed. This was accomplished through individual review of the radiology reports. Because the goal of the study was to determine downstream effects of incidental thyroid nodule detection, the re-identification of a known finding disqualified patients from further review. Example language that conveyed known thyroid findings were “stable thyroid nodule” or “unchanged from previous exam.”
After culling the list down to only the “apparently new thyroid incidentaloma” cases, each case was subjected to a detailed review of the medical record to classify patients as “notified” or “not notified” of the incidental finding, as per the requirements of the modified study design. Acceptable forms of data to classify a patient as “notified” were documentation of communication of the finding to the patient in an office visit or telephone note, or evidence of subsequent thyroid directed testing, such as a laboratory test, imaging study, biopsy, or thyroid surgery.
Patients who were classified as “notified” were contacted by telephone for a structured survey. The survey took place between 10 and 23 months after the imaging study of interest was performed. They were asked what they knew about why the test on which their thyroid incidentaloma was identified had been ordered, what they understood the results to be, and how they received those results. Patients were also asked to complete the six-item short form of the State–Trait Anxiety Inventory (STAI) for the time around the thyroid finding. This validated, widely used survey instrument was developed by Spielberger in 1962 for quantifying anxiety as both a stable personality trait and a state related to specific events. It was subsequently modified into the shortened form, which we utilized, by Marteau and Bekker (11).
The relational database for the registry was created in Filemaker Pro v12 (Santa Clara, CA). Descriptive statistics were calculated using Stata v12 (College Station, TX).
During the one-year study period from July 1, 2011, to June 30, 2012, there were 109,220 imaging studies performed on which a thyroid incidentaloma could have been detected (67% chest x-ray, 21% CT scans of the neck or chest, 4% MRI of the neck, 5% PET scan, and 2% myocardial perfusion scan). Only 146 studies documented such a finding (<1%). This reporting rate is much lower than published prevalence rates of 16% for incidental thyroid nodules on CT and MRI studies (12,13), even considering the large numbers of films that were chest x-rays, which raised the concern that radiologist reporting of thyroid nodules was incomplete.
Upon detailed review, 21 of these cases were excluded because the test was actually thyroid directed (n=7), the record was a duplicate (n=12), the patient did not speak English (n=1), or the patient was younger than 18 years of age (n=1). This left 125 patients eligible for inclusion in the study (Fig. 1).
The imaging study on which the thyroid finding was identified was a CT scan of the chest, abdomen, and pelvis in 47/129 cases (38%), a CT scan of the chest alone in 37 cases (30%), a PET scan in 20 cases (16%), a neck CT scan in 11 cases (9%), a MRI scan in 9 cases (8%), and a chest x-ray in 1 case (0.8%). Women comprised 61% of the cases (76/125).
The indication for imaging was cancer care (e.g., workup, staging, or restaging) in 72/125 (57%) cases; a chest process (e.g., lung disease re-evaluation or pulmonary embolus evaluation) in 16/125 (13%) cases; an abdominal process (e.g., aortic aneurysm evaluation or abdominal pain) in 10/125 (8%) cases; trauma evaluation in 10/125 (8%) cases; evaluation of a neck or facial mass (e.g., lipoma or parotid gland abnormality) in 8/125 (6%) cases; workup of a neurologic or musculoskeletal problem (e.g., weakness, voice change or neck pain) in 7/125 (6%) cases; and follow-up of a different incidentaloma in 2/125 (2%) cases.
Among the 125 patients in the initial case listing, 46 had a thyroid finding that was not new based on manual review of the radiology report. This left 79 patients with apparently new thyroid nodules—one of the criteria for inclusion in the registry.
Because the approved study design would only allow contact with patients who had clearly been notified of their finding, the next step was to exclude those who had not been notified of their finding. Among the 79 patients with new thyroid nodules, nine were apparently appropriately not notified by their care provider about the finding, which is consistent with thyroid nodule management guidelines in place at the time (14,15). Their nodule was <1cm and did not have a worrisome appearance. Thus, further workup was not necessarily indicated. Another 35 patients appeared not to have been notified of the finding by their care provider for unknown reasons, or they were notified but no notation was made in the chart to indicate the finding was communicated to the patient. This brought the number of patients eligible for the telephone survey portion of the registry development down to less than half: 35/79 (44%). This loss of potentially eligible patients raises concerns about bias in the sample.
The 35 patients classified as “notified” were contacted from a Dartmouth Hitchcock Medical Center phone number for the telephone survey portion of the registry development. More than half were unreachable (20/35 patients) after five attempts at different times of the day and days of the week—the phone rang continuously with no answer, was out of service, or people overtly declined participation when reached.
Of the 15 patients successfully contacted, six did not recall being told about their thyroid finding, and so could not complete all survey questions (most questions required knowledge that one had a thyroid finding). At least two of these six patients with no recollection of their thyroid finding saw an endocrinologist, as documented in the medical record, and one of these two patients underwent needle biopsy. The other three patients in this group were referred to endocrinologists outside the Dartmouth Hitchcock hospital system, according to the medical record, but records of those visits were not available, making it impossible to know whether the patient actually saw an endocrinologist.
Among the nine patients who were able to complete the survey in full, management was widely variable. Three had an ultrasound, and no biopsy was determined to be necessary (nodule size range 1.2–1.6cm), two had ultrasounds and biopsies (nodule size range 1–3cm, cytology benign in both instances), two underwent total thyroidectomy (for 3.1 and >5cm nodules), and two had biopsy recommended but the patients declined them (for 0.9 and 5.5cm nodules).
None of the nine patients who were able to complete the telephone survey in full self-identified their finding as “incidental.” While a clinician might recognize such a finding as potentially not clinically important, the study participants did not make this distinction. When asked about the results of their test on which the thyroid finding was uncovered, a typical response was “they found a cancer” or “they found something wrong with my thyroid gland.” Their answers to the questions on the STAI thus reflected the levels of concern one might expect with any potentially cancerous finding. The STAI range of responses allowed for each item are “not at all,” “somewhat,” “moderately so,” and “very much so.” For the item “I felt upset,” three people said they were “not at all” upset by the finding, five “moderately so,” and one “very much so.” For the item “I was worried,” one person said they were “not at all worried,” four “somewhat,” one “moderately so,” and two “very much so.”
The natural history of incidentalomas is poorly understood, a major problem for patients, clinicians, researchers, and the healthcare system. Patients identified as having incidentalomas and their providers are faced with a dilemma about how aggressive to be with workup and subsequent treatment, given that the finding is not causing symptoms and has been present for an unknown length of time at the time of identification. It is very hard to predict how much of a threat the finding is to the patient. For researchers, the validity of clinical trial data are threatened if incidentally detected cancers are included in studies because the clinical behavior of incidentalomas may be more indolent than those that were identified as a result of symptoms, potentially falsely improving trial results intended to be applied to cancers that were found because of symptoms. Last, the healthcare system is faced with a potentially overwhelming time and financial cost for managing findings that may not be of clinical importance. Incidentalomas distract personnel from clinical problems of potentially greater threat to their patients' lives on the individual level, and on a broader level take up workforce time that could be better applied to medical problems more germane to the population as a whole.
The goal of this study was to create a registry that would begin to provide key answers needed to alleviate these questions. What is the natural history of such findings, and which ones need treatment? How are patients affected by incidentalomas from a healthcare and personal perspective? Are some people and their findings better off managed right away, while other people and findings are better watched over time? The registry proved difficult to execute from a design standpoint for both logistical and ethical reasons. The observational registry design originally outlined could not be operationalized as hoped, and even after modification, there were numerous threats to data validity. Marked variation was found in radiologist and clinician reporting and management of such findings, a widespread problem as previously reported by Hoang et al. (16).
Calculation of annual case estimates of incidentalomas is difficult. While >100,000 eligible imaging studies were done at the authors' institution in the year under review, only 125 patients were identified by radiologists as having a thyroid nodule in their radiology reports (<1% prevalence). Autopsy studies show about 50% of clinically normal thyroids have a nodule by pathologic examination (3). On directed review of CT scans and MRIs that include the thyroid gland, the identifiable prevalence is lower—but still approximately 16%—many fold higher than the <1% reporting rate in this study (13,17). Northern New England, where the study was performed, is a very low-intensity practice area (18), and the radiology department culture is one in which incidentalomas not deemed to be clinically relevant are not routinely reported. It should also be noted that the radiology department at the authors' institution does not hold the vascular ultrasound studies in its database—these are done in a separate department. This would reasonably be expected to affect the incidentaloma detection rate, since carotid ultrasounds are a known mechanism by which thyroid nodules might be detected (19). However, the findings of this study suggest that the identification rate at institutions across the country is in all likelihood quite variable, and might change year to year as institutional culture changes, which would make it hard to achieve a registry aim of accurate incidence estimates over time and across institutions.
One solution to improve annual case estimate reliability would be to mandate reporting of every thyroid incidentaloma, but this may result in more harm than good if it increases patients' exposure to potentially unnecessary procedures and if it results in increased healthcare costs related to the findings. Without knowing the natural history of incidentalomas, it would not be possible to determine if procedures and costs resulting from mandated reporting were necessary or helpful.
Patient recollection of incidental findings, even when clearly notified, is variable. Nearly half of the patients who apparently were aware of their thyroid finding based on medical record documentation and episodes of downstream testing had no recollection of it when they were contacted. This finding is important information for investigators who may wish to do similar research—patients may have entire episodes of care about which they have a limited ability to converse outside the immediate context of the care experience.
Patient self-conception of a finding will affect their reports of anxiety associated with it. Patients contacted as part of this study uniformly did not conceptually self-identify their thyroid findings as incidental (i.e., “found by accident”). This has important ramifications for questions related to anxiety or stress created by the finding. If a person presumes that all findings on an imaging study represent an imminent threat to life, the meaning of the anxiety and stress ratings on a standardized scale is qualitatively quite different than if the person is informed that the finding may or may not be of clinical significance, and then asked to quantify their anxiety and stress. In that case, the anxiety would likely be due to the uncertainty of clinical significance, as opposed to the anxiety of a threat to life. Both findings might result in stress and anxiety but for different reasons. This suggests that further work in this area would require careful stratification of groups by self-assessment of clinical relevance to that particular person based on their own values.
The IRB did not allow any patient with an incidentaloma that had not clearly been previously notified, as demonstrated through documentation in the medical record, to be contacted, stating that the risk was too high that patients would learn they had an incidentaloma but had not been notified. This might lead to anxiety among participants in the study and could potentially undermine the physician–patient relationship. This was a reasonable concern with which the authors could not disagree. However, the consequence was that it was not possible to obtain full data on the downstream effects on the patient and the healthcare system for more than half of the patients. This made it impossible to achieve a key goal of the registry: to obtain accurate estimates of downstream costs—both personal and financial—of thyroid incidentaloma detection.
One solution would be to query providers about the decision to notify or not notify a patient of their finding, but this approach has problems. It was not always clear which physician was responsible for reporting incidental finding results to a patient. In addition, it was not clear that answers obtained about reasons for notification/non-notification of patients would be remembered or reported accurately. Lastly, querying providers may change provider behavior—potentially inappropriately changing activity related to thyroid nodules. Thus, we deemed this potential solution unacceptable.
Embedding an incidentaloma registry in an intervention designed to monitor and promote appropriate incidentaloma management is recommended. In this design, every incidentaloma of certain qualities would be required to be reported in the radiology report. The initial bar for intervention on the finding would necessarily be set “low” so that only findings that all informed stakeholders agreed were low risk would be observed. Patients with low-risk findings deemed suitable for observation but who decided upon intervention would not be stopped from doing so, but would be monitored via the registry for outcomes related to their interventions. With the recent publication by the American College of Radiology about suggested management strategies for incidentally detected thyroid nodules, there is now a reasonable place from which to start such a registry (20). The major previous white paper from this organization on abdominal incidentalomas has been adopted widely based on recent survey data (21), so it is likely that the thyroid white paper would be acceptable to radiology colleagues as a starting point in this work. In both white papers, algorithms for management suggest potential actions and follow-up strategies for various types of findings, although they do not indicate specific stopping points when “monitoring” is suggested as an action—a shortcoming of the papers. This intervention/registry design would also likely be acceptable to an IRB because it would potentially decrease risks patients are currently exposed to by variable management of incidentalomas. Consent would be integrated into the consent for the procedure itself, with the option to opt out. The registry would undergo data monitoring at regularly specified intervals using a variation on the concept of a data safety monitoring board for clinical trials. The bar for intervention would gradually be moved up or down systematically as indicated based on the interim data on the benefits and harms associated with the incidentaloma observation and intervention.
While the early days of such an intervention/registry might employ management recommendations that potentially result in overtreatment for some patients, over time, the management recommendations could become more conservative or aggressive as knowledge of natural history accumulates about typical behaviors of incidentalomas, protecting future patients from over- or under-treatment, and the medical system from the burden of unnecessary care.
While some would argue that this design would take too long or not be equitable to all patients, the situation we are in now is hardly better. The identification of incidental thyroid nodules and intervention rates surrounding it currently depend on whether the radiologist reports the nodule, whether the physician responsible for the test interpretation notifies the patient, and then the local cultures and practices around the management of such findings in general. This variability is worse overall for patients than a systematic approach that addresses and attempts to solve the problem on a population level.
An intervention/registry design would also be advantageous because it would result in more uniform reporting by radiologists, if decision rules at the institutional level were created about what to report. This design would also create a reasonable expectation of patient notification and a better chance to educate patients about the meaning of such findings, allowing improved estimates of the patient experience of incidentaloma detection and improved education of the public about the pros and cons of both intervention and observation of incidentalomas. And, it would provide a clear way to disseminate current best practices for incidentaloma workup and follow-up testing with ordering providers, which would improve quality of care overall. Last, and most importantly, it would promote appropriate and consistent management of incidentalomas, which is the ultimate goal of this research endeavor.
The findings should provide insight for investigators interested in developing an observational registry of incidentalomas of the thyroid or other organs. The challenges outlined from the authors' experience demonstrate major negative ramifications for incidentaloma registry enrollment, threatening both data completeness and validity. Several solutions that might typically be employed in a registry to solve the problems identified have the potential to worsen incidentaloma detection, potentially placing patients in harm's way from both under- and overtreatment. The interventional/observational design proposed here may provide the basis for an alternative path forward for future researchers.
This study was supported by National Cancer Institute grant 5UC2CA148259—Comparative Effectiveness of Advanced Imaging in Cancer, and the Department of Veterans Affairs. This work was presented in an earlier draft form as a poster at the Preventing Overdiagnosis Conference, Hanover, NH, September 2013. The views expressed do not necessarily represent the views of the Department of Veterans Affairs or the United States Government.
All authors declare that we have no commercial financial conflict of interest.