A Tabulated Summary of the FDG PET Literature

Search Criteria Analysis
The literature search was performed using the databases Medline/Healthstar 1993 - 2000 and Biosis Previews 1993 - 2000 for articles and abstracts published from January 1993 - June 2000. All key word combinations, including FDG PET, PET, and specific oncologic, neurologic, and cardiac applications were searched. Printed copies of The Journal of Clinical Positron Imaging (1998 - 2000) and The Journal of Nuclear Medicine abstracts (1996 - 2000) also were used. Only articles/abstracts in English were used, with the exception of a few English abstracts of non-English-language articles that provided complete information. Both dedicated PET and newer low-cost PET technology (for example, coincidence imaging) studies were included.

The only exceptions to our search time period occurred in the neurological and cardiac application categories. Specifically for myocardial applications, the Medline search extended back to 1986, with a focus on literature assessing viable myocardium. For dementia and seizure workup, the Medline search extended back to 1980, with respective foci on literature assessing accuracy in diagnosing individual patients with dementia and on literature assessing PET performance with respect to evaluating potential candidates for neurosurgery.

All literature that was not clear with respect to methods and/or reporting was excluded. Furthermore, any article/abstract that reported on a study with five or fewer individuals also was excluded. A total of 775 articles/abstracts were retrieved from the literature for our review. Approximately 8 articles could not be obtained from interlibrary requests to outside libraries. The data analysis used 473 unique articles/abstracts (specifically 151 abstracts and 322 articles), and 302 were excluded as per the inclusion/exclusion criteria. The spreadsheets listed a total of 561 article/abstract entries, of which 17 were repeated across several spreadsheets to which they were applicable and 71 were repeated within spreadsheets in multiple applications.

Inclusion Criteria
(1) Abstracts and articles reporting data within which sensitivity (sens), specificity (spec), positive predictive value (ppv), negative predictive value (npv), accuracy (acc), and management change (mgmt) values were either partially or fully listed or could be partially or fully derived for FDG PET imaging in the 22 different oncologic areas, cardiac viability area, and dementia and seizure work-up areas. In addition, some studies (e.g., seizure) were listed with FDG PET contributions to clinical issues without accompanying accuracy data. Only data with stated or derived total patient studies or total lesions were incorporated into the weighted averages. In those instances in which CT data were found in the PET literature satisfying the inclusion criteria, these were also listed.

(2) Oncologic studies drawn from the period January 1993 - June 2000; dementia and seizure studies from January 1980 - June 2000; and cardiac studies from January 1986 - June 2000.

(3) Response-to-treatment articles were included in the spreadsheets where a 2 x 2 table could be created from the reported data for: responders/nonresponders versus increased FDG/decreased FDG. In those instances in which a 2 x 2 table could not be formulated, the article was excluded.

Note that three articles were also included that provided no numerical information about FDG PET accuracy but had some useful features, which are described in the comments field. These articles, therefore, have no bearing on the weighted averages summarizing all the literature data. These studies by Bischoff et al. (46), Holthoff et al. (197), and Rozental et al. (354) were all part of the monitoring response application.

Exclusion Criteria
(1) Case reports, review/tutorial articles, and studies with 5 or fewer patients.

(2) Articles not in English. However, abstracts in English of articles not in English but with relevant information were included.

Data Analysis
Data analysis was performed using simple weighted averages. Therefore, studies with more patients were weighted more than studies with fewer patients to arrive at estimates of the sensitivity, specificity, and, when possible, management changes. Weighting is the easiest method to use on such a large number of studies, each of which may or may not present a full 2 x 2 table of outcomes. No attempt was made to perform a formal meta-analysis.

In instances in which articles and abstracts included data for multiple categories (e.g., diagnosis/staging/recurrence), the entire article entry was listed in each of the three individual categories (diagnosis, staging, and recurrence) to preserve the entirety of a study´s reporting and to represent that study´s contribution to data for that category for both this report and possible future analyses that might be looking for all references including data for a given category (e.g., specifically for recurrence.) Only data relevant to a specific category was used in the weighted average for that category (e.g., in calculating the weighted average in the recurrence category, only the recurrence portion of the article´s data was used, even though data for diagnosis also may have been listed).

The number in the total patient studies column sometimes exceeded that in the total number of patients column for a given entry line (e.g., in instances in which patients may have had multiple FDG PET scans). For each line entry of data, the total patient studies or total lesions were listed upon which the 2 x 2 table was based for calculating a given line of data (e.g., if 58 patients had 62 scans from which the true positive (TP), true negative (TN), false positive (FP), and false negative (FN) values were counted, 62 was listed for total patient studies).

In those instances in which articles/abstracts had data broken down for various reported subgroups (e.g., mediastinal and hilar lymph nodes or lymph nodes <1 cm), total patient studies for each subgroup would be listed (as explained above), but often these subgroups would have overlapping patients. In terms of the data analysis, when a given study provided overall values in addition to listing various subgroup values, the overall value was used in the weighted average. When an overall value was not listed, the subgroup data was weighted in by the total patient studies value from which it was generated (or by total lesions, if listed by lesions). The only exceptions occurred in the lung cancer spreadsheet/staging section in the four articles by Baum et al. (36), Tatsumi et al. (424), Ryu et al. (356), and Marom et al. (290). When these studies reported subgroup values for the full patient study count multiple times, the subgroup values were averaged and weighted into the weighted average formula by the total patient studies for one group only.

Back to top.