The review addressed a clear question and inclusion criteria were defined. Details on reference standard were lacking. A limited search was conducted. The review was restricted to published studies, so relevant studies (published and unpublished) may have been missed. Appropriate steps were taken to minimise bias and error when extracting data and assessing quality; it was unclear whether such steps were taken when selecting studies for inclusion. Study quality was assessed using relevant criteria. Details on the included studies (other than results of the quality assessment) were not reported and so it was not possible to determine the generalisability of the review findings.
A limited synthesis was conducted. The main pooling of results was based on simple summing of the data in the 2x2 tables. Pooling data in this way had potential to produce biased summary estimates; it would have been preferable to use more robust models (such as the bivariate random-effects model). There was substantial heterogeneity across studies and this was not adequately investigated. Some of the tabular and graphical displays lacked clarity: it was unclear whether they referred to PET and/or PET/CT studies and it was unclear how many studies contributed to pooled analyses.
The summary estimates did not support the authors' conclusions of high accuracy of PET and PET/CT with sensitivity of only 74%, which combined with limitations in the analysis and potential for missing studies means that the authors' conclusions should be interpreted with caution.