| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Nuclear Medicine |
1 From the Department of Radiology/Division of Nuclear Medicine, Molecular Imaging Program at Stanford (MIPS), Stanford University Medical Center, 300 Pasteur Dr, H-0101, Stanford, CA 94305. From the 2004 RSNA Annual Meeting. Received December 22, 2005; revision requested February 20, 2006; revision received May 4; accepted June 1; final version accepted August 10. Address correspondence to A.Q. (e-mail: aquon{at}stanford.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT MATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
Materials and Methods: The study conformed to HIPAA standards, and the need for informed consent was waived by the institutional review board that approved the study. FDG PET/CT reports of 712 patients were reviewed to identify patients with malignant bone lesions. Fifty-nine patients (30 female and 29 male patients; age range, 10–82 years) with 113 lesions were analyzed. With use of confirmation from histopathologic examination or clinical follow-up, the PPVs of the integrated examination and of the stand-alone CT and PET components of the examination were calculated. The results were stratified according to cancer type, chemotherapy status, and number of bone lesions and were compared by using Fisher exact tests.
Results: Of 47 lesions with positive findings at both PET and CT, 46 were malignant and one was benign, for a PPV of 98%. Of 31 lesions with positive findings at PET and negative findings at CT, 19 were malignant and 12 were benign, for a PPV of 61%. Of 35 lesions with negative findings at PET and positive findings at CT, six were malignant and 29 were benign, for a PPV of 17%. Independently, the PPV of all lesions with positive findings at PET was significantly higher than that of all lesions with positive findings at CT. Chemotherapy status for lesions with positive findings at CT and the number of lesions per patient had a statistically significant effect on the PPV of examinations (P = .02 and P < .001, respectively).
Conclusion: PET/CT has a very high PPV for bone metastases (98%) when the findings at PET and CT are concordant; however, in lesions with discordant PET and CT findings at the integrated examination, PPV is markedly diminished.
© RSNA, 2007
Fluorine 18 fluorodeoxyglucose (FDG) positron emission tomography (PET) is a well-established modality for standard oncologic staging, restaging, and treatment monitoring evaluations (1–3). A key issue that is less well studied is the performance of FDG PET in accurately depicting bone metastases that would potentially have a large effect on patient treatment. Although radionuclide bone scanning with technetium 99m (99mTc) methyl diphosphonate has been the standard means of evaluating individuals suspected of having bone metastases, FDG PET may be comparable in accuracy, depending on the tumor type (4–9). More recently, integrated PET/computed tomography (CT) has revealed various implications for evaluating bone metastases (10–13). First, integrated PET/CT can help better differentiate whether FDG-avid lesions are truly located within bone versus adjacent soft tissue (14). Second, the CT data are a potentially valuable addition to the FDG PET information.
Although bone lesions that are suspicious for malignancy at both the PET and CT portions of a PET/CT examination are likely to be malignant, lesions that are deemed positive at one portion of the examination but appear benign at the other are potentially problematic. Few data exist for describing how often discordant PET and CT results occur and the predictive value of the examination when these discordant lesions exist. Therefore, the purpose of our study was to retrospectively evaluate the positive predictive value (PPV) of FDG PET/CT in the identification of malignant bone lesions when the findings at PET and CT are directly discordant and concordant.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
|
PET/CT Scanning and Image Interpretation
All patients that met our study criteria had undergone intravenous injection of 12–15 mCi (444–555 MBq) of FDG. After injection of FDG, patients waited 45–90 minutes before undergoing scanning. Plasma glucose measurements were obtained in all patients and ranged from 76 to 210 mg/dL (4.2–11.6 mmol/L) (allowable range at our institution is 70–250 mg/dL [3.9–13.9 mmol/L]). All patients were scanned with a PET/CT unit (Discovery LS; GE Medical Systems, Waukesha, Wis). The parameters for CT image acquisition were as follows: An initial scout view was obtained with 30 mAs and 120 kVp, followed by spiral CT at 0.8 second per rotation with 100 mAs, 149 kVp, section thickness of 5 mm, and a 4.25-mm interval. Intravenous contrast material was not administered. PET emission images were obtained with a weight-based protocol, with 4–6 minutes of acquisition time per bed position. All PET images were reconstructed by using an iterative algorithm, with CT-based attenuation correction applied.
Image interpretation had been individually performed by four board-certified nuclear medicine physicians (two authors, A.Q. and S.S.G.) with more than 30 collective years of PET and PET/CT experience (range, 6–10 years) and six radiologists (one author, R.J.H.) with more than 45 collective years of body imaging experience (range, 3–28 years). A separate dictated report was generated for each examination, and no joint report was produced. All data from this study were based on the original interpretation, and no consensus reinterpretation of the scans was performed. The interpreting radiologist had read all CT images with a viewing station (Centricity RA1000; GE Medical Systems) by using transverse sections. The nuclear medicine physician used a workstation (Xeleris 1.1; GE Medical Systems) for viewing all PET images, using the transverse, coronal, and sagittal views.
At the original image interpretation, the following PET criteria were used in the diagnosis of a positive lesion: evaluation of the signal intensity of foci on the basis of semiquantitative standardized uptake value measurements (greater than 2.0); visual assessment of signal intensity by means of comparison to physiologic structures such as the liver, bowel, and background soft tissue; and visual assessment of the focality and location of each lesion. In cases of increased FDG activity in such benign conditions as marrow expansion and degenerative disease, interpreting physicians attempted to account for potentially false-positive results by more heavily considering other factors such as the focality and location of each lesion rather than merely the standardized uptake value. At CT, typical criteria used for making the diagnosis of a malignant bone lesion included identification of focal-appearing lytic or sclerotic lesions. Additional features that favored malignancy included marrow replacement, soft-tissue component, endosteal scalloping, cortical breakthrough, periosteal reaction, expansile appearance, or associated pathologic fracture.
Follow-up
For benign lesions, clinical follow-up was the primary means of confirmation (reference standard). Specifically, this required that the patient remained free of symptoms and/or that the lesions remained stable and appeared benign at subsequent imaging for at least 6 months.
For malignant lesions, assessment of malignancy was based on results of histopathologic examination, additional interval imaging studies, and clinical follow-up. Lesions with histopathologic confirmation (13 of 113 lesions total) were considered malignant without any further correlative data. When a tissue-based diagnosis was not available, lesions were considered malignant when at least one of four possible criteria were met, as follows: (a) Lesion progression was demonstrated at two subsequent imaging examinations (bone scanning, magnetic resonance [MR] imaging, dedicated CT, or additional integrated PET/CT) (75 of 113 lesions), (b) lesion progression was demonstrated with a single subsequent imaging examination and at clinical follow-up (seven of 113 lesions), (c) bone disease was found and documented at clinical evaluation and physical examination (five of 113 lesions), and (d) lesions were positive at the initial PET or CT examination and then regressed after treatment with chemotherapy (13 of 113 lesions).
Statistical Analysis
In our core group analysis, we calculated the PPV for correctly identifying malignant bone lesions in several different scenarios that can occur: (a) lesions that are positive at both the PET and CT portions of the examination, (b) lesions that are positive at PET and negative at CT, and (c) lesions that are negative at PET and positive at CT. Furthermore, we evaluated the stand-alone PPV of PET and CT if each were used alone.
A calculation of the sensitivity of the PET/CT examination was not performed because our initial retrospective review included only lesions that were detected with either PET or CT, and therefore the true total number of bone lesions in our total population of 712 patients is unknown.
In addition, we also analyzed the effect of several modifying factors that could affect the PPV of PET/CT, including the number of lesions per patient, the chemotherapy status of the patient, and the tumor type. The results were stratified according to cancer type, chemotherapy status, and number of bone lesions and were compared by using Fisher exact tests.
To account for the effect of multiple lesions per patient, a subject-specific generalized estimating equations logistic regression was fit to the data, with lesion malignancy as the dependent variable; detection type (CT alone, PET alone, or PET/CT), chemotherapy (yes or no), tumor type, and number of lesions as independent variables; and patient as the clustering variable. A P value of .05 was considered indicative of a statistically significant difference. Furthermore, in addition to a lesion-by-lesion analysis, we investigated the PPV of the examinations on a patient-by-patient basis. For this analysis, the patient was considered to be correctly identified as having metastatic bone disease if at least one of the identified suspicious lesions was in fact malignant. In other words, we calculated the PPV of the examination in addressing the question, "Does a patient with a positive test result have any metastatic bone disease?" All statistical analyses were done with statistical software (Stata 9.1; Stata, College Station, Tex).
The 11 patients with diffuse bone metastases (at least 10 lesions each) were analyzed separately (see below). The reason for this was twofold: First, the large number of lesions within this small group of patients would have skewed the PPV results upward, and, second, in clinical practice these patients do not present the same diagnostic challenge as do patients with a small number of suspicious bone lesions. Therefore, in these patients with diffuse disease, we simply performed a patient-by-patient analysis rather than a lesion-by-lesion analysis.
|
RESULTS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
Concordant and Discordant Findings
Of the 47 lesions identified as positive for malignancy at both the PET and CT portions of the integrated examination, 46 proved to be malignant and one was benign. Therefore, the PPV for bone lesions that were positive at both PET and CT was 98% (95% confidence interval [CI]: 88%, 100%).
The PET and CT findings were discordant for 66 of the 113 bone lesions (58%). In the 31 lesions that were positive at the PET and negative at the CT portions of the integrated examination, 19 proved to be malignant and 12 were benign, for a PPV of 61% (95% CI: 42%, 78%) (Fig 1). Of the 35 lesions that were negative at PET and positive at CT, six were malignant and 29 were benign, for a PPV of 17% (95% CI: 7%, 33%) (Fig 2).
|
|
|
|
Modifying Factors
The PPV changed depending on the number of lesions identified per patient (Table 2). In the patients with a single abnormal focus at the CT portion of the examination (21 patients, 21 lesions identified with CT alone, eight true-positive lesions), the PPV of CT alone was 38% (95% CI: 18%, 62%). The PPV of CT-positive foci increased to 70% (95% CI: 51%, 84%) for patients with two to four abnormal foci (19 patients, 33 lesions identified, 23 true-positive lesions) and to 75% (95% CI: 55%, 89%) for patients with five to nine foci (six patients, 28 lesions identified, 21 true-positive lesions).
|
The PPV of the integrated PET/CT examination was 43% (17 of 40) when a patient had a solitary bone focus seen with either CT or PET (but not both). For patients with a single focus that was concordantly abnormal (suspicious for malignancy) at both PET and CT (eight patients, eight lesions, seven true-positive lesions), the PPV was 88% (95% CI: 47%, 100%; P = .04 vs lesions diagnosed as positive with CT alone and P = .09 vs lesions positive with PET alone). The PPV for concordant examinations was 100% (95% CI: 86%, 100%; P = .004 vs lesions positive with CT alone and P = .029 vs lesions positive with PET alone) for patients with two to four PET/CT foci (12 patients, 21 lesions, 21 true-positive lesions) and 100% (95% CI: 85%, 100%; P = .03 vs lesions positive with CT alone and P > .99 vs lesions positive with PET alone) for patients with five to nine PET/CT foci (five patients, 18 lesions, 18 true-positive lesions) (Table 2).
On a patient-by-patient basis, the overall PPVs changed only modestly (Table 2), and the overall conclusions are similar to those with the lesion-by-lesion analysis described earlier. For patients with only one suspicious bone malignancy at examination, a positive result at stand-alone PET or CT has a low predictive value. Independent confirmation of these lesions is warranted. However, for patients with more than one lesion identified with any of the methods, the chances are much higher that the patient has at least some level of malignant bone disease.
We also analyzed the effect of chemotherapy on the PPV. The effect of chemotherapy was substantial for lesions identified at CT. Overall, the PPV of lesions detected with CT alone was 63% (95% CI: 52%, 74%) for all patients in our study. The PPV of lesions identified at CT was 76% (95% CI: 61%, 88%) for the 27 patients who did not undergo chemotherapy and 50% (95% CI: 34%, 66%) for the 32 patients who had undergone chemotherapy at some point before their examination (P = .02) (Table 3). We attempted to compare the PPVs of the different examination modalities according to tumor type. In general, lesions that were positive at both PET and CT had a higher PPV across tumor type than did lesions identified with only one modality (Fig 3). In addition, lesions identified as positive at PET and negative at CT usually had a higher PPV across tumor types than did lesions that were negative at PET and positive at CT. However, because of the small sample sizes when data were analyzed separately according to tumor type, these differences were not statistically significant.
|
|
Diffuse Disease
Of the 11 patients with diffuse disease, diffuse, active bone metastases was confirmed in 10 patients with clinical follow-up and imaging. Parotid cancer was diagnosed in one patient with diffuse sclerotic lesions at CT. The lesions proved to be residual lesions subsequent to treatment with chemotherapy. Of the 10 patients with demonstrated malignancies, two patients had false-negative findings at CT (Hodgkin lymphoma, non-Hodgkin lymphoma) and one patient had a false-negative finding at PET (breast cancer). The other seven patients had positive findings at both CT and PET. All but one of the patients with diffuse disease had undergone chemotherapy prior to the examinations (the exception being a patient with tumor of unknown primary origin with bone lesions positive at both PET and CT).
|
DISCUSSION |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
In the discordant cases, the PPV of a lesion with positive findings at PET and negative findings at CT was 61%. In contrast, bone lesions that were negative at PET and positive at CT had a PPV of 17%, which suggests that FDG PET is a better predictor of bone metastases than CT when the examinations are discordant. But perhaps more surprisingly, when our data are compared with those from earlier studies on the performance of FDG PET used alone for bone metastases (15,16), we can infer that PET is not a very strong predictor of a true bone metastasis when the associated CT finding is negative, particularly for solitary foci. Our results appear to corroborate more recent data (17) that suggest that the coregistered CT image may improve the PPV of the overall PET/CT examination. From these data, it appears necessary to obtain additional correlative evidence (from additional imaging examinations, physical examination, and/or biopsy) to confirm that a solitary bone focus that is negative at CT and positive at PET is truly malignant. More specifically, lesions that were positive at only one component of the PET/CT examination had PPVs that were notably lower than those with the stand-alone modality. This is relevant because many physicians may use the PPV, sensitivity, and specificity data of the stand-alone modalities to guide treatment decisions (15).
We also attempted to find patterns of improved and diminished predictive value depending on the number of lesions identified per patient, chemotherapy status, and the type of tumor involved. The PPV of suspicious lesions increased with an increasing number of lesions identified per patient. For patients with just one suspicious lesion, only lesions that were positive at both the PET and CT portions of the examination had a high likelihood of being malignant. For patients with more than four lesions per examination, however, the PPV of PET-positive lesions was quite high (regardless of whether the lesions were also identified at CT). This suggests that the corroborative benefit of an integrated PET/CT examination is greatest with those patients for whom only one lesion or a few lesions are identified. As larger numbers of lesions are identified, most or all of the lesions identified with PET can, with some confidence, be considered malignant.
Chemotherapy status did not affect the PPV of PET examinations. CT examinations in patients who had never undergone chemotherapy, however, had a significantly higher PPV than did CT examinations in patients with prior chemotherapy exposure. This emphasizes the importance of understanding patients' chemotherapy history when evaluating results from integrated PET/CT examinations.
Our study has limitations. First, histopathologic confirmation of malignancy was uncommon. Although this is not surprising, it limits our reference standard to other imaging modalities and clinical follow-up. In terms of tumor types, breast and prostate cancers were most notably underrepresented and blood-borne malignancies overrepresented. This is a reflection of the patient population at our institution and is one of the limitations of a single-institution study. In addition, the type of data we present would be more useful if broken down by tumor type. We did not have adequate numbers of patients to meaningfully perform these subgroup analyses. Finally, the CT portion of the examinations was performed without contrast material. Although this might affect the evaluation of soft-tissue metastases, it is unlikely to have a substantial influence on the evaluation of bone metastases. Future comparisons should include an evaluation of FDG PET versus 99mTc methyl diphosphonate bone scanning and perhaps a newer-generation PET tracer such as fluoride ion.
In conclusion, the PPV of integrated PET/CT imaging in the evaluation of bone malignancy is very high (98%) when the two portions of the examination are in agreement. However, the PET and CT examinations appear to have discordant findings relatively frequently. When the examinations are discordant, PET is far more accurate than CT in the characterization of bone lesions (PPV, 61% for PET vs 17% for CT; negative predictive value, 83% for PET vs 39% for CT) but is clearly not as accurate as when the two examinations are concordant. Furthermore, in patients with solitary bone lesions for which the PET and CT findings are discordant, the PPV for integrated PET/CT is even lower (43%) than that in patients with multiple lesions. Thus, an additional adjunctive examination (eg, MR imaging or biopsy) may be necessary.
|
ADVANCES IN KNOWLEDGE |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Abbreviations: CI = confidence interval • FDG = fluorine 18 fluorodeoxyglucose • PPV = positive predictive value
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, all authors; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, all authors; clinical studies, all authors; statistical analysis, A.V.T., S.S.G., A.Q.; and manuscript editing, all authors
|
References |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  F. Breibach, A. Julian, C. Laurent, E. Anglade, A. Constantin, N. Sans, J.-L. Tricoire, P. Brousset, M.-B. Delisle, and A. Gomez-Brouchet Contribution of the 2-[18F]-fluoro-2-deoxy-D-glucose positron emission tomography/computed tomography to the diagnosis of primary osseous Hodgkin lymphoma BMJ Case Reports, May 25, 2009; 2009(may25_1): bcr0820080696 - bcr0820080696. [Abstract] [Full Text]
|
|
|
|
 |
|
 S. Carkaci, H. A. Macapinlac, M. Cristofanilli, O. Mawlawi, E. Rohren, A. M. Gonzalez Angulo, S. Dawood, E. Resetkova, H. T. Le-Petross, and W.-T. Yang Retrospective Study of 18F-FDG PET/CT in the Diagnosis of Inflammatory Breast Cancer: Preliminary Data J. Nucl. Med., February 1, 2009; 50(2): 231 - 238. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Du, I. Cullum, T. M. Illidge, and P. J. Ell Fusion of Metabolic Function and Morphology: Sequential [18F]Fluorodeoxyglucose Positron-Emission Tomography/Computed Tomography Studies Yield New Insights Into the Natural History of Bone Metastases in Breast Cancer J. Clin. Oncol., August 10, 2007; 25(23): 3440 - 3447. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
