Test Performance of Positron Emission Tomography and Computed Tomography for Mediastinal Staging in Patients with Non–Small-Cell Lung Cancer

A Meta-Analysis

View larger version (28K): [in a new window]   Figure 1. Reports evaluated for inclusion in the meta-analysis. The initial search took place from 1966 through 1 June 2002, and the supplemental search took place from 1998 through 27 March 2003. PET = positron emission tomography.

View larger version (30K): [in a new window]   Figure 2. Individual study estimates of sensitivity and 1 – specificity of computed tomography for identifying mediastinal metastasis. Error bars represent 95% CIs. Three studies reported results by using both the patient and lymph nodes or lymph node stations as the units of analysis; these 3 studies are listed twice (70, 74, 75).

View larger version (27K): [in a new window]   Figure 3. Individual study estimates of sensitivity and 1 – specificity of positron emission tomography with 18-fluorodeoxyglucose for identifying mediastinal metastasis. Error bars represent 95% CIs. Five studies reported results by using both the patient and lymph nodes or lymph node stations as the units of analysis; these 5 studies are listed twice (70, 73-75, 96).

View larger version (23K): [in a new window]   Figure 4. Summary receiver-operating characteristic curves and 95% CIs for mediastinal staging with positron emission tomography with 18-fluorodeoxyglucose (FDG-PET) and computed tomography (CT). Individual study estimates of sensitivity and 1 – specificity are shown for FDG-PET (open circles) and CT (s). The approximate points on the curves where FDG-PET and CT operate in current practice are indicated (solid circle and solid square, respectively).

View larger version (16K): [in a new window]   Figure 5. Summary receiver-operating characteristic curves for mediastinal staging with positron emission tomography with 18-fluorodeoxyglucose in patients with and without mediastinal lymph node enlargement on computed tomography (CT). Individual study estimates of sensitivity and 1 – specificity are shown for positron emission tomography with 18-fluorodeoxyglucose in patients with enlarged lymph nodes (s) and without enlarged lymph nodes (open circles). The 2 receiver-operating characteristic curves are nearly identical. However, in patients with enlarged lymph nodes on CT, studies tend to cluster on a portion of the curve at which sensitivity is favored over specificity. In patients without lymph node enlargement, studies tend to cluster on a portion of the curve at which specificity is favored over sensitivity. The approximate points on the curves where positron emission tomography with 18-fluorodeoxyglucose operates in current practice in patients with and without lymph node enlargement are indicated (solid square and solid circle, respectively). The discriminant function that separates the 2 groups of patients is shown (dashed line) (P = 0.002 by nonparametric permutation test).

View larger version (24K): [in a new window]   Figure 6. Post-test probabilities of mediastinal metastasis after computed tomography (CT) and positron emission tomography with 18-fluorodeoxyglucose (FDG-PET). Post-test probabilities are shown as a function of pretest probability in patients with positive FDG-PET results and enlarged lymph nodes on CT (circles), patients with positive FDG-PET results and no enlarged lymph nodes on CT (squares), patients with negative FDG-PET results and enlarged lymph nodes on CT (triangles), and patients with negative FDG-PET results and no enlarged lymph nodes on CT (diamonds).

View larger version (34K): [in a new window]   Appendix Figure. Post-test probabilities of mediastinal metastasis after computed tomography (CT) and positron emission tomography with 18-fluorodeoxyglucose (FDG-PET). Post-test probabilities are shown as a function of pretest probability in patients with positive FDG-PET results and enlarged lymph nodes on CT (solid circles), patients with positive FDG-PET results and no enlarged lymph nodes on CT (solid squares), patients with negative FDG-PET results and enlarged lymph nodes on CT (solid triangles), and patients with negative FDG-PET results and no enlarged lymph nodes on CT (solid diamonds). When unconditional estimates of FDG-PET performance are used to make the calculations, post-test probabilities are overestimated when FDG-PET results are positive and CT shows enlarged lymph nodes (open circles), underestimated when FDG-PET results are positive and CT shows no enlarged lymph nodes (s), overestimated when FDG-PET results are negative and CT shows enlarged lymph nodes (s), and underestimated when FDG-PET results are negative and CT shows no enlarged lymph nodes (open diamonds).