Aims and background: A meta-analysis of diagnostic test performance was conducted to compare the results of relevant studies reporting diagnostic accuracy values for mediastinal staging in patients with non-small cell lung cancer (NSCLC). This paper deals with the two most accurate imaging techniques currently in use: positron emission tomography (PET) with FDG and computed tomography (CT). A statistical pooling method was used to perform a quantitative meta-analysis aimed at demonstrating the potential advantage of one of these two methods.
Methods: Studies in all languages published between 1998 and 2005 that examined the use of FDG-PET and CT for mediastinal staging in NSCLC patients, enrolled at least 18 participants, and provided enough data to allow calculation of sensitivity and specificity rates were considered eligible for the quantitative meta-analysis. Statistical methods to pool the overall estimates of sensitivity and specificity and to compare the discriminant power of PET and CT were discussed and used.
Results: Of the 13 studies included in the analysis, 12 reported greater accuracy of FDG-PET than CT in detecting mediastinal lymph node metastases. The sensitivity of FDG-PET ranged from 50% to 100%. The estimate of the overall sensitivity was 0.83% with 95% CI (0.749-0.913). Specificity ranged from 79% to 100%, with an overall estimated specificity of 0.87% with 95% CI (0.80-0.95). For CT, the sensitivity and specificity ranged from 50% to 97% and 58% to 94%, respectively; the overall estimate was 0.68% with 95% CI (0.582-0.788) and 0.76% with 95% CI (0.668-0.859). The summary receiver operating characteristic (SROC) approach was used to assess the superior diagnostic accuracy of one of the two methods. The areas under the two SROC curves were AUC(PET) = 0.909 vs. AUC(CT) = 0.794.
Conclusions: Numerical and visual results of the meta-analysis of recent relevant reports agreed that FDG-PET is more accurate than CT in identifying mediastinal lymph node metastases in non-small cell lung cancer.