TO THE EDITOR:
We wish to comment on certain aspects of an interesting recent paper by Hoffman et al. (1).
In initial evaluation of new patients who might be suffering from one of the dementias, it is common practice in many institutions to obtain an MRI scan of the head, largely to rule out various organic lesions such as stroke or neoplasms; furthermore, occasionally changes in the hippocampi and inferior medial temporal cortical regions may indicate early Alzheimer’s disease (AD). However, in several AD clinics, MRI has been supplemented by a second examination intended to evaluate certain aspects of brain physiologic function. Hoffman et al. (1) make a case for the use of 18F-FDG PET. In assessing alternatives to PET involving radionuclide imaging, Hoffman et al. consider several studies performed with SPECT and a variety of regional cerebral blood flow (rCBF) tracers. One of these was the first report from our group on histopathologically confirmed prospective SPECT diagnoses in 18 patients, in which we examined certain features of rCBF SPECT for its efficacy in the diagnosis of AD (2). Unfortunately, the authors missed our following report (3), in which we described our results in 54 patients, also with histopathologic confirmation. To assist readers in evaluating the role of SPECT in the study of patients with AD and other dementias, we thought that it might be useful to review our more recent SPECT results.
We performed SPECT rCBF scans on 504 patients with possible dementia and 32 elderly healthy volunteers (age range, 56–89). The first half of our experience involved the use of inhaled 133Xe and an appropriate 4-detector scanner, while the most recent 254 studies were done with 99mTc-hexamethylpropyleneamine oxime and a triple-camera scanner. Histopathologic correlation with prospective diagnoses for the presence or absence of AD was obtained in 79 patients, with the following results: sensitivity = 53/61 (86.9%; 95% confidence limits [CL], 75.2%–93.8%); specificity = 13/18 (72.2%; CL, 46.4%–89.3%); positive predictive value = 53/59 (89.8%; CL, 78.5%–95.8%); and negative predictive value = 13/20 (65.0%; CL, 40.9%–83.7%).
Of the 79 diagnoses, 77 were made at autopsy, 1 was made by both biopsy and subsequent autopsy, and 1 by biopsy alone. We compared the sensitivity of each of the tracers for the diagnosis of AD and find that they do not differ significantly (P = 0.29, 2-tailed Fisher exact test) (3).
In our series, the interval between initial prospective diagnosis and histopathologic correlation had a mean of approximately 48 mo, which is to be expected when patient survival occasionally exceeds 10 y after initial SPECT study. Hoffman et al. (1) are fortunate in having a somewhat shorter interval of 30.8 mo, but theirs is a small series of 22 patients; with an increasing number of patients, the mean length of time will increase as well, as will the extent of disease at autopsy and their chances of making false-negative diagnoses.
At the present time, our AD center’s clinical staff has become adept at diagnosing typical AD even in its early stages; in these cases rCBF SPECT scans are no longer obtained. However, the staff frequently refers for SPECT patients whose symptoms and test results are confusing. The result often suggests one of the frontotemporal entities, which will become an important distinction if a specific treatment for AD becomes available. Lewy body disease, which is usually combined with AD, may also present a perplexing clinical aspect; 11 such cases have been referred for SPECT scans, and 10 have yielded true-positive SPECT results for AD. The presence of Lewy body disease may contribute an additional diagnostic sign in the form of abnormally reduced occipital cortical rCBF (4), which may occur in the absence of occipital cortical Lewy bodies at autopsy. However, one must keep in mind that reduced occipital cortical flow may also be seen in certain patients with depression but no dementia (5).
On the basis of our experience, we believe that because it is widely available and is less expensive than PET, rCBF SPECT, which performs well as a diagnostic test, might be combined with MRI as the experimental approach of choice for patients suspected of dementia whose clinical diagnosis is unclear.
ACKNOWLEDGMENTS
This work was supported in part by National Institutes of Health grant 2-P30-AG 12300-(01–07) and by Nycomed Amersham Inc.
REFERENCES
REPLY:
We appreciate the letter of Dr. Bonte et al. regarding our article (1). As the authors note, we did not cite their more recent publication (2), but we did acknowledge their first and very important study evaluating the use of brain blood flow in dementias using SPECT (3). Because of the limit on the number of references that could be cited, we chose to use the earlier publication by the authors because it was one of the first studies to use functional imaging techniques and histologic verification rather than clinical diagnosis in the evaluation of dementia. Dr. Bonte and his colleagues should be recognized for their contributions and pioneering work in the assessment of dementia with functional imaging techniques and their correlation of imaging results with pathologic diagnoses rather than clinical diagnoses.
Dr. Bonte and his colleagues now provide data on 79 patients who have had both pathologic verification of diagnosis and SPECT regional cerebral blood flow (rCBF) imaging. These new data confirm the results of their previous work with SPECT, showing a sensitivity for the diagnosis of Alzheimer’s disease (AD) at approximately 87%, a specificity of 72%, and a calculated accuracy of 83.5%. These values are very similar to the values we determined for 18F-FDG PET in our publication.
An important fact to keep in mind is the purpose of our study. We were specifically interested in confirming, with pathological verification, the data in the literature that bilateral temporoparietal hypometabolism using FDG PET is the metabolic abnormality associated with Alzheimer’s disease. We also wanted to determine that the sensitivity, specificity, and diagnostic accuracy of the metabolic pattern of bilateral temporoparietal hypometabolism allows for differentiation between AD and other degenerative causes of dementia. These major goals of our study were satisfied.
Both articles by Bonte et al. (2,3) as well as our article (1) note the difficulties of evaluating dementia with metabolic and functional assessments, particularly regarding confounding pathologic diagnoses. The coexistence of Lewy body disease, vascular disease, or other degenerative processes with AD clearly is one source of confusing diagnosis that must be considered. This possibility for a coexisting disease was another concern in our study and part of why a unilateral reduction in FDG uptake was not considered as a hallmark for the diagnosis of AD. Unilateral reductions may well favor the metabolic correlate of a remote vascular event or a disease process other than AD.
There are also differences in approaches to image interpretation and analysis between our study and those of Bonte et al., which might possibly affect sensitivity, specificity, and accuracy calculations. In their more recent article, the sensitivity for visual interpretation was not specifically given for the 54 patients in the study; however, imaging sensitivity using a combined visual and semiquantitative approach was 37/43 (86%) and the specificity was 8/11 (73%). The sensitivity results are essentially the same between their earlier and later studies, which is important in light of the increased number of patients. In their earlier report, both bilateral and unilateral perfusion changes were interpreted as positive for AD. In 37 of their true-positive cases, 22 were based on bilateral reductions in perfusion (22 true-positives and zero false-positives). In 15 of the 37 true-positive cases the AD diagnosis was based on a unilateral reduction (15 true-positives but 3 false-positives). These data confirm our belief that bilateral reductions in either metabolism or perfusion will provide a more accurate test. When unilateral reductions are used the specificity and false-positive rate will increase. In our study we chose to use visual criteria to grade the images rather than semiquantitative techniques because visual analysis is more commonly used in a clinical setting. Such differences in approach make it difficult to directly compare sensitivity, specificity, and diagnostic accuracy. Nevertheless the results appear to be similar for the SPECT rCBF and FDG PET approaches.
We do not agree with the statement that early diagnosis and differentiation of AD by clinical criteria is easy. Numerous studies from the literature give varying rates of diagnostic accuracy for clinical evaluation. In our own series of diagnostically challenging patients, the sensitivity, specificity, and diagnostic accuracy for the clinical diagnosis of AD was lower than for image-based diagnosis. It appears that at Dr. Bonte’s institution the clinical and image-based evaluation of patients with dementia are now more consistent with the approach used in our study. In particular, SPECT rCBF imaging is used in those patients with confusing symptoms and test results. This approach is consistent with the evolving general practice of the medical community as well as at many academically oriented memory disorder or dementia clinics. However, functional imaging assessments may become more important in the early evaluation of patients with mild memory loss or difficult-to-characterize early dementia, or diagnostically challenging patients where a more accurate diagnosis is needed to confirm the suspected diagnosis. This early, accurate diagnosis is becoming much more important because new drug therapies for AD are now becoming available.
Our results and those of Bonte et al. appear to be very similar as to sensitivity, specificity, and diagnostic accuracy, and reflect the potential power of using a functional imaging technique in evaluating patients with dementia.