A Tabulated Summary of the FDG PET Literature

Because the default analysis (analysis 1) of our 22 spreadsheets listed all the literature values used that included multiple listings of abstracts/articles both within spreadsheets (e.g., both in the diagnosis/staging and staging applications of a particular cancer) and across spreadsheets, some overlapping of patient studies occurred. To analyze the effects of listing the broad view of the literature values used we performed 5 additional data analyses to study the effects on PET sens/spec/acc by selecting out certain studies according to the following criteria:

(1) Analysis 1 (default analysis): Included all literature used that had some overlap of total patient studies within and across spreadsheets.

(2) Analysis 2: Used data from only those studies that reported jointly the sensitivity and specificity values for PET and PET accuracy values that included one other PET statistical ratio column (e.g., reported PET accuracy and also PET sensitivity).

(3) Analysis 3: Repeated the default analysis, including data from only full research articles. All abstracts were specifically excluded.

(4) Analysis 4: Repeated analysis 2, including data from only full research articles. All abstracts were excluded.

(5) Analysis 5: Pooled sens/spec/acc values for PET across all available studies that provided TP/TN/FP/FN values for each cancer and for all cancers together. This is a formal pooling analysis, using data from each study to construct a large 2 x 2 table. Note that all four cells (TP/TN/FP/FN) must be available, that is, information for patients both with and without disease is required. Therefore, some studies used in some of the weighted averaging subanalysis formulations (e.g., reporting sens only) could not be included in the pooling subanalysis, and thus slight discrepancies exist between the article subsets used in the weighted averages and those used in the pooling.

(6) Analysis 6: Repeated the default analysis excluding all data from nondedicated PET machines (e.g., coincidence imaging).

Results are provided in Table 25.

Back to top.