Twenty-one studies were included in the review (1,217 patients, 1,391 adrenal lesions). All studies included an appropriate patient spectrum and avoided incorporation bias. One study fulfilled all criteria. Less than 20% of studies used an appropriate reference standard and less than 20% avoided verification bias. Blinding of was poorly reported.
Sensitivity ranged from 76% to 100% and specificity ranged from 73% to 100%. Summary sensitivity was 97% (95% CI 93% to 98%) and summary specificity was 91% (95% CI 87% to 94%). There was substantial heterogeneity in both sensitivity and specificity estimates (I2=88%). The summary positive likelihood ratio was 11.1 (95% CI 7.5 to 16.3) and the summary negative likelihood ratio was 0.04 (95% CI 0.02 to 0.8). The AUC was 0.98. Summary estimates did not vary depending on method of image interpretation. The two variables to show an association with sensitivity in the regression analysis (p<0.05) were prospective design and full verification. Multi-observer interpretation, withdrawals, clinical data, blinding of the index test and clinical population were associated with specificity (p<0.05). At pre-test probabilities of 5%, 25% and 50% there were no significant differences in the post-test probability of a negative test results between the subgroups. A positive PET result increased the pre-test probabilities of 5%, 25% and 50% to 67.2%, 92.9% and 97.5% for patients with an adrenal mass with no known cancer. The authors stated that there was no evidence of publication bias.