Erik Supratik Mittra MD, PhD, Abstract Co-Author: Nothing to Disclose

Andrei Iagaru MD, Presenter: Research Consultant, MediGene AG

Sanjiv S. Gambhir MD, PhD, Abstract Co-Author: Board of Directors, Lumera Corporation Stockholder, Lumera Corporation Stockholder, Pfizer Inc Consultant, Spectrum Dynamics Ltd Stockholder, Spectrum Dynamics Ltd Grant, Johnson & Johnson Committee member, Amgen Inc Scientific Advisory Board, Novartis AG Scientific Advisory Board, Siemens AG Royalties, Reed Elsevier Scientific Advisory Board, Genentech, Inc Scientific Advisory Board, General Electric Company Grant, General Electric Company Research collaboration, GlaxoSmithKline plc Scientific Advisory Board, GlaxoSmithKline plc Scientific Advisory Board, Intronn Inc Research collaboration, Intronn Inc Grant, Intronn Inc Scientific Advisory Board, Lumen Therapeutics Consultant, MediGene AG Scientific Advisory Board, MediGene AG Consultant, Millennium Pharmaceuticals, Inc Research collaboration, Pfizer Inc Grant, Pfizer Inc Consultant, Koninklijke Philips Electronics NV Scientific Advisory Board, Koninklijke Philips Electronics NV Consultant, Pathwork Diagnostics Grant, Bayer AG Speaker, Siemens AG Scientific Advisory Board, Varian Medical Systems, Inc Scientific Advisory Board, VisualSonics Inc

The metabolic information in PET (i.e., SUV) and anatomic information in CT (i.e., tumor size/volume) are related but distinct. However, for varied reasons, the abnormality on CT is not always present or readily identifiable. The available literature describing the relationship between PET and CT-based tumor burden assessment is limited and dated. Therefore, we were prompted to evaluate this issue more comprehensively.

103 distinct lesions were measured across 3 tumor types (Lymphoma: 43 lesions, 10 patients; Lung cancer: 30 lesions, 11 patients; Head and Neck cancer: 30 lesions,12 patients). The most recent PET/CT scans with those malignancies were randomly selected. Up to 5 lesions per scan were measured in two dimensions, first on PET, then on CT, by the same individual. Significance of the difference between the long-axis, short-axis, and 2D "volume" measurements between PET and CT were made by a two-tailed T-test (p < 0.05). Pearson product moment correlation coefficients with their respective p-values were also calculated between the SUV value of the lesion and the square of the difference between the PET and CT measurements.

There were no significant differences found between those measurements taken from PET versus CT (all p > 0.05). This was true for the long-axis, short-axis, and "volume" measurements, as well as for all 3 tumor types. For all 103 lesions, the mean size on PET was 2.03 cm, 1.44 cm, and 4.74 cm2 versus 2.06 cm, 1.43 cm, and 4.41 cm2 on CT for the long-axis, short-axis, and "volume" measures, respectively. Additionally, correlation of the SUV values versus the square of the difference between the PET and CT measurements shows a modest but significant positive relationship for both lymphoma and head and neck cancer (r = 0.36, p = 0.02; and r = 0.53, p = 0.001; respectively). No such relationship was found for lung cancer (r = 0.00, p = 0.99).  

Measurement of tumor size based on the metabolic information from PET is not significantly different from that based on the anatomic information from CT. In lymphoma and head and neck cancer, the more intense the uptake, the greater was the (minimal) measurement error. This was not true for lung cancer.

Across varied tumor types, measurement of tumor size by PET alone is not significantly different from CT, such that PET can be used to assess tumor burden when the CT is not available or clear.

Mittra, E, Iagaru, A, Gambhir, S, Accuracy of Tumor Measurements by 18F-FDG PET.  Radiological Society of North America 2009 Scientific Assembly and Annual Meeting, November 29 - December 4, 2009 ,Chicago IL. http://archive.rsna.org/2009/8013872.html