FDG-PET Performs Poorly in Lung Cancer Diagnosis

CHICAGO – The diagnostic accuracy of FDG-PET in lung cancer performed below levels described in previous reports and varied widely among U.S. centers in a secondary analysis of a large phase III clinical trial.

"FDG-PET performed poorly for diagnosing non–small cell lung cancer in a national sample of clinical stage 1 patients," Dr. Eric L. Grogan said at the annnual meeting of the of the American Society of Clinical Oncology.

The current National Comprehensive Cancer Network guidelines recommend the use of fluorodeoxyglucose positron emission tomography (FDG-PET) for the diagnosis of NSCLC based on studies showing a high degree of accuracy for this diagnostic tool, notably a sensitivity of 94% and a median specificity of 83% in a meta-analysis of 40 studies (JAMA 2001;285:914-24).

Others have reported, however, that FDG-PET performs poorly at single institutions in regions of endemic fungal lung diseases (Ann. Thor. Surg. 2011;92:428-32 and Lung Cancer 2002;36:297-301), observed Dr. Grogan, of Vanderbilt-Ingram Cancer Center in Nashville, Tenn.

Among 682 patients in the American College of Surgeons Oncology Group (ACOSOG) Z4031 trial, the overall accuracy of FDG-PET was 73%, the sensitivity 82%, and the specificity only 31%.

The series is the largest to date evaluating the accuracy of FDG-PET in patients with known or suspected clinical stage 1 NSCLC. In addition, it is generalizable to clinical practice because multiple FDG-PET scanners were used and the scans were performed in community and academic centers and interpreted by multiple radiologists, Dr. Grogan said.

"Results of PET scans in this population should be interpreted cautiously, and reasons for the poor test performance should be explored in other studies," he said.

Discussant Dr. Tetsuya Mitsudomi, who is the chief of thoracic surgery at Aichi Cancer Center Hospital in Nagoya, Japan, said FDG-PET shows reasonable sensitivity, but very low specificity compared with previous studies.

"I think this reflects the real world," he said. "So, the lung cancer diagnosis cannot be made on the basis of PET positivity alone."

Investigators at 51 sites in 39 cities enrolled 969 patients with known or clinically suspicious stage 1 lesions between 2004 and 2006 to evaluate the value of proteomic analysis in diagnosing NSCLC (the results were presented at ASCO 2010). FDG-PET scans were available for 682 patients. All underwent surgical resection. Analyses were performed for all patients and for sites with more than 25 patients.

PET avidity was determined by the radiologist’s description of lesion activity or by the reported maximum standard uptake value (SUV). Avidity was classified in four categories: category 1 was no avidity/not cancer (SUV = 0), category 2 was low avidity/not likely cancer (SUV 0 to less than 2.5), category 3 was avidity/possibly cancer (SUV 2.5 to less than 5.0), and category 4 was high avidity/likely cancer (SUV 5.0 or more).

Among the 682 patients, there were 566 cancers and 116 benign cases. In all, 82% of the cancerous lesions were PET avid, and "surprisingly, 69% of the benign lesions were avid," Dr. Grogan said.

Patients with cancer were significantly older (67 vs. 61 years; P less than .001) and had larger lesions (26 mm vs. 20 mm; P less than .001).

The positive predictive value of FDG-PET was 85% and negative predictive value 26%. This translates into 80 false positives and 101 false negatives. The majority of false positives were found to be granulomas (69%), he observed. Eleven of the false negatives were 10 mm or less.

Not surprising, FDG-PET accuracy improved with lesion size, Dr. Grogan said. The accuracy was less than 50% for lesions less than 20 mm, but greater than 80% for lesions larger than 30 mm. "Above 30 mm, the accuracy did not seem to improve," he observed.

In the eight cities that had more than 25 patients, the sensitivity varied significantly, from a low of 67% in Los Angeles to a high of 91% in Durham, N.C. (P = .03), Dr. Grogan said, without explanation. Specificity ranged from 15% in Birmingham, Ala., to 46% in Philadelphia, but this did not reach statistical significance because of the small number of benign cases at each institution (P = .72).

Dr. Mitsudomi said he could not explain the reason for the heterogeneity, especially in terms of the specificity, between centers.

"It’s not possible to remove all the false positives if you use FDG, but newer tracers are being developed and they may increase the specificity rate," he added.

Dr. Grogan reported no disclosures. Dr. Mitsudomi reported having a consulting/advisory role with Boehringer Ingelheim, Kyowa Hakko Kirin, Lilly, and Pfizer, and receiving honoraria from AstraZeneca, Chugai Pharma, Lilly, and Roche.☐

CHICAGO – The diagnostic accuracy of FDG-PET in lung cancer performed below levels described in previous reports and varied widely among U.S. centers in a secondary analysis of a large phase III clinical trial.

"FDG-PET performed poorly for diagnosing non–small cell lung cancer in a national sample of clinical stage 1 patients," Dr. Eric L. Grogan said at the annnual meeting of the of the American Society of Clinical Oncology.

The current National Comprehensive Cancer Network guidelines recommend the use of fluorodeoxyglucose positron emission tomography (FDG-PET) for the diagnosis of NSCLC based on studies showing a high degree of accuracy for this diagnostic tool, notably a sensitivity of 94% and a median specificity of 83% in a meta-analysis of 40 studies (JAMA 2001;285:914-24).

Others have reported, however, that FDG-PET performs poorly at single institutions in regions of endemic fungal lung diseases (Ann. Thor. Surg. 2011;92:428-32 and Lung Cancer 2002;36:297-301), observed Dr. Grogan, of Vanderbilt-Ingram Cancer Center in Nashville, Tenn.

Among 682 patients in the American College of Surgeons Oncology Group (ACOSOG) Z4031 trial, the overall accuracy of FDG-PET was 73%, the sensitivity 82%, and the specificity only 31%.

The series is the largest to date evaluating the accuracy of FDG-PET in patients with known or suspected clinical stage 1 NSCLC. In addition, it is generalizable to clinical practice because multiple FDG-PET scanners were used and the scans were performed in community and academic centers and interpreted by multiple radiologists, Dr. Grogan said.

"Results of PET scans in this population should be interpreted cautiously, and reasons for the poor test performance should be explored in other studies," he said.

Discussant Dr. Tetsuya Mitsudomi, who is the chief of thoracic surgery at Aichi Cancer Center Hospital in Nagoya, Japan, said FDG-PET shows reasonable sensitivity, but very low specificity compared with previous studies.

"I think this reflects the real world," he said. "So, the lung cancer diagnosis cannot be made on the basis of PET positivity alone."

Investigators at 51 sites in 39 cities enrolled 969 patients with known or clinically suspicious stage 1 lesions between 2004 and 2006 to evaluate the value of proteomic analysis in diagnosing NSCLC (the results were presented at ASCO 2010). FDG-PET scans were available for 682 patients. All underwent surgical resection. Analyses were performed for all patients and for sites with more than 25 patients.

PET avidity was determined by the radiologist’s description of lesion activity or by the reported maximum standard uptake value (SUV). Avidity was classified in four categories: category 1 was no avidity/not cancer (SUV = 0), category 2 was low avidity/not likely cancer (SUV 0 to less than 2.5), category 3 was avidity/possibly cancer (SUV 2.5 to less than 5.0), and category 4 was high avidity/likely cancer (SUV 5.0 or more).

Among the 682 patients, there were 566 cancers and 116 benign cases. In all, 82% of the cancerous lesions were PET avid, and "surprisingly, 69% of the benign lesions were avid," Dr. Grogan said.

Patients with cancer were significantly older (67 vs. 61 years; P less than .001) and had larger lesions (26 mm vs. 20 mm; P less than .001).

The positive predictive value of FDG-PET was 85% and negative predictive value 26%. This translates into 80 false positives and 101 false negatives. The majority of false positives were found to be granulomas (69%), he observed. Eleven of the false negatives were 10 mm or less.

Not surprising, FDG-PET accuracy improved with lesion size, Dr. Grogan said. The accuracy was less than 50% for lesions less than 20 mm, but greater than 80% for lesions larger than 30 mm. "Above 30 mm, the accuracy did not seem to improve," he observed.

In the eight cities that had more than 25 patients, the sensitivity varied significantly, from a low of 67% in Los Angeles to a high of 91% in Durham, N.C. (P = .03), Dr. Grogan said, without explanation. Specificity ranged from 15% in Birmingham, Ala., to 46% in Philadelphia, but this did not reach statistical significance because of the small number of benign cases at each institution (P = .72).

Dr. Mitsudomi said he could not explain the reason for the heterogeneity, especially in terms of the specificity, between centers.

"It’s not possible to remove all the false positives if you use FDG, but newer tracers are being developed and they may increase the specificity rate," he added.

Dr. Grogan reported no disclosures. Dr. Mitsudomi reported having a consulting/advisory role with Boehringer Ingelheim, Kyowa Hakko Kirin, Lilly, and Pfizer, and receiving honoraria from AstraZeneca, Chugai Pharma, Lilly, and Roche.☐

CHICAGO – The diagnostic accuracy of FDG-PET in lung cancer performed below levels described in previous reports and varied widely among U.S. centers in a secondary analysis of a large phase III clinical trial.

"FDG-PET performed poorly for diagnosing non–small cell lung cancer in a national sample of clinical stage 1 patients," Dr. Eric L. Grogan said at the annnual meeting of the of the American Society of Clinical Oncology.

The current National Comprehensive Cancer Network guidelines recommend the use of fluorodeoxyglucose positron emission tomography (FDG-PET) for the diagnosis of NSCLC based on studies showing a high degree of accuracy for this diagnostic tool, notably a sensitivity of 94% and a median specificity of 83% in a meta-analysis of 40 studies (JAMA 2001;285:914-24).

Others have reported, however, that FDG-PET performs poorly at single institutions in regions of endemic fungal lung diseases (Ann. Thor. Surg. 2011;92:428-32 and Lung Cancer 2002;36:297-301), observed Dr. Grogan, of Vanderbilt-Ingram Cancer Center in Nashville, Tenn.

Among 682 patients in the American College of Surgeons Oncology Group (ACOSOG) Z4031 trial, the overall accuracy of FDG-PET was 73%, the sensitivity 82%, and the specificity only 31%.

The series is the largest to date evaluating the accuracy of FDG-PET in patients with known or suspected clinical stage 1 NSCLC. In addition, it is generalizable to clinical practice because multiple FDG-PET scanners were used and the scans were performed in community and academic centers and interpreted by multiple radiologists, Dr. Grogan said.

"Results of PET scans in this population should be interpreted cautiously, and reasons for the poor test performance should be explored in other studies," he said.

Discussant Dr. Tetsuya Mitsudomi, who is the chief of thoracic surgery at Aichi Cancer Center Hospital in Nagoya, Japan, said FDG-PET shows reasonable sensitivity, but very low specificity compared with previous studies.

"I think this reflects the real world," he said. "So, the lung cancer diagnosis cannot be made on the basis of PET positivity alone."

Investigators at 51 sites in 39 cities enrolled 969 patients with known or clinically suspicious stage 1 lesions between 2004 and 2006 to evaluate the value of proteomic analysis in diagnosing NSCLC (the results were presented at ASCO 2010). FDG-PET scans were available for 682 patients. All underwent surgical resection. Analyses were performed for all patients and for sites with more than 25 patients.

PET avidity was determined by the radiologist’s description of lesion activity or by the reported maximum standard uptake value (SUV). Avidity was classified in four categories: category 1 was no avidity/not cancer (SUV = 0), category 2 was low avidity/not likely cancer (SUV 0 to less than 2.5), category 3 was avidity/possibly cancer (SUV 2.5 to less than 5.0), and category 4 was high avidity/likely cancer (SUV 5.0 or more).

Among the 682 patients, there were 566 cancers and 116 benign cases. In all, 82% of the cancerous lesions were PET avid, and "surprisingly, 69% of the benign lesions were avid," Dr. Grogan said.

Patients with cancer were significantly older (67 vs. 61 years; P less than .001) and had larger lesions (26 mm vs. 20 mm; P less than .001).

The positive predictive value of FDG-PET was 85% and negative predictive value 26%. This translates into 80 false positives and 101 false negatives. The majority of false positives were found to be granulomas (69%), he observed. Eleven of the false negatives were 10 mm or less.

Not surprising, FDG-PET accuracy improved with lesion size, Dr. Grogan said. The accuracy was less than 50% for lesions less than 20 mm, but greater than 80% for lesions larger than 30 mm. "Above 30 mm, the accuracy did not seem to improve," he observed.

In the eight cities that had more than 25 patients, the sensitivity varied significantly, from a low of 67% in Los Angeles to a high of 91% in Durham, N.C. (P = .03), Dr. Grogan said, without explanation. Specificity ranged from 15% in Birmingham, Ala., to 46% in Philadelphia, but this did not reach statistical significance because of the small number of benign cases at each institution (P = .72).

Dr. Mitsudomi said he could not explain the reason for the heterogeneity, especially in terms of the specificity, between centers.

"It’s not possible to remove all the false positives if you use FDG, but newer tracers are being developed and they may increase the specificity rate," he added.

Dr. Grogan reported no disclosures. Dr. Mitsudomi reported having a consulting/advisory role with Boehringer Ingelheim, Kyowa Hakko Kirin, Lilly, and Pfizer, and receiving honoraria from AstraZeneca, Chugai Pharma, Lilly, and Roche.☐