Imaging for Colorectal Liver Metasases - A Metaanalysis
Abstract purpose: To perform a meta-analysis to obtain on a per-patient and per-lesion basis sensitivity estimates of state of the art computed tomography (CT), magnetic resonance (MR) imaging, fluorine-18-fluorodeoxyglucose (FDG) positron emission tomography (PET) and PET/CT for the detection of colorectal liver metastases.
Materials and method: The MEDLINE and EMBASE databases were searched for relevant original articles published from January 2004 to October 2009 on CT, MR imaging, FDG-PET and PET/CT for the detection and characterization of colorectal liver metastases.
Criteria for inclusion were: prospective study; study population >10 patients; histopathologically proven primary colorectal carcinoma; diagnostic imaging was performed to identify and characterize liver lesions; intra-operative findings, histopathology and/or follow-up were used as reference standard; contingency table could be constructed. Raw data was documented and sensitivity and specificity estimates were calculated on a per-patient basis for the four different modalities. Sensitivity estimates were calculated on a per-lesion basis. All analyses were based on nonlinear mixed-effects approach.
Results: The internet-search yielded a total of 5890 articles, of which 24 fulfilled all inclusion criteria. The included studies evaluated 2613 patients (57% male, 43% female) with a mean age of 62.1 years (range; 23–93 years). Sensitivity estimates on a per-patient basis for helical CT, 1.5-T MR imaging, FDG-PET and PET/CT were 80.7 % (95 % CI; 60.4–91.9), 67.1 % (95 % CI; 59.5–73.9), 95.4 % (95 % CI; 93.2–97.0) and 96.8 (95 % CI; 94.6–98.2) respectively (Fig. 1); FDG-PET and PET/CT were the most accurate modalities; differences were significant (all p < 0.05) compared to helical CT and MR imaging. On a per-patient basis, specificity estimates for helical CT, 1.5-T MR imaging, FDG-PET and PET/CT were 88.7 % (95 % CI; 50.5–98.4), 92.9 % (95 % CI; 89.9–95.1), 91.7 % (95 % CI; 86.0–95.2) and 96.2 % (95 % CI; 91.3–98.4) (Fig. 2). Differences between the modalities were not significant (all p > 0.05). On a per-lesion basis, sensitivity estimates for helical CT, 1.5-T MR imaging, FDG-PET and PET/CT were 70.0 % (95 % CI; 65.5–74.2), 87.3 % (95 % CI; 82.0–91.2), 75.2 % (95 % CI; 58.6–86.7) and 58.0 % (95 % CI; 47.1–68.2), respectively (Fig. 3). MR imaging had the highest sensitivity; differences compared to the other modalities were significant (all p < 0.05).
(Enlarge Image)
Figure 1.
Per-patient sensitivity with 95 % confidence intervals.
(Enlarge Image)
Figure 2.
Per-patient specificity with 95 % confidence intervals.
(Enlarge Image)
Figure 3.
Per-lesion sensitivity with 95 % confidence intervals.
Conclusion: FDG-PET and PET/CT have the highest sensitivity estimates on a per-patient basis, but MR imaging has the highest sensitivity on a per-lesion basis. All sensitivity estimates, except for MR imaging, were lower on a per-lesion basis compared to the sensitivity estimates on a per-patient basis. Specificity estimates on a per-patient basis were comparable between the four modalities.
Abstract and Introduction
Introduction
Abstract purpose: To perform a meta-analysis to obtain on a per-patient and per-lesion basis sensitivity estimates of state of the art computed tomography (CT), magnetic resonance (MR) imaging, fluorine-18-fluorodeoxyglucose (FDG) positron emission tomography (PET) and PET/CT for the detection of colorectal liver metastases.
Materials and method: The MEDLINE and EMBASE databases were searched for relevant original articles published from January 2004 to October 2009 on CT, MR imaging, FDG-PET and PET/CT for the detection and characterization of colorectal liver metastases.
Criteria for inclusion were: prospective study; study population >10 patients; histopathologically proven primary colorectal carcinoma; diagnostic imaging was performed to identify and characterize liver lesions; intra-operative findings, histopathology and/or follow-up were used as reference standard; contingency table could be constructed. Raw data was documented and sensitivity and specificity estimates were calculated on a per-patient basis for the four different modalities. Sensitivity estimates were calculated on a per-lesion basis. All analyses were based on nonlinear mixed-effects approach.
Results: The internet-search yielded a total of 5890 articles, of which 24 fulfilled all inclusion criteria. The included studies evaluated 2613 patients (57% male, 43% female) with a mean age of 62.1 years (range; 23–93 years). Sensitivity estimates on a per-patient basis for helical CT, 1.5-T MR imaging, FDG-PET and PET/CT were 80.7 % (95 % CI; 60.4–91.9), 67.1 % (95 % CI; 59.5–73.9), 95.4 % (95 % CI; 93.2–97.0) and 96.8 (95 % CI; 94.6–98.2) respectively (Fig. 1); FDG-PET and PET/CT were the most accurate modalities; differences were significant (all p < 0.05) compared to helical CT and MR imaging. On a per-patient basis, specificity estimates for helical CT, 1.5-T MR imaging, FDG-PET and PET/CT were 88.7 % (95 % CI; 50.5–98.4), 92.9 % (95 % CI; 89.9–95.1), 91.7 % (95 % CI; 86.0–95.2) and 96.2 % (95 % CI; 91.3–98.4) (Fig. 2). Differences between the modalities were not significant (all p > 0.05). On a per-lesion basis, sensitivity estimates for helical CT, 1.5-T MR imaging, FDG-PET and PET/CT were 70.0 % (95 % CI; 65.5–74.2), 87.3 % (95 % CI; 82.0–91.2), 75.2 % (95 % CI; 58.6–86.7) and 58.0 % (95 % CI; 47.1–68.2), respectively (Fig. 3). MR imaging had the highest sensitivity; differences compared to the other modalities were significant (all p < 0.05).
(Enlarge Image)
Figure 1.
Per-patient sensitivity with 95 % confidence intervals.
(Enlarge Image)
Figure 2.
Per-patient specificity with 95 % confidence intervals.
(Enlarge Image)
Figure 3.
Per-lesion sensitivity with 95 % confidence intervals.
Conclusion: FDG-PET and PET/CT have the highest sensitivity estimates on a per-patient basis, but MR imaging has the highest sensitivity on a per-lesion basis. All sensitivity estimates, except for MR imaging, were lower on a per-lesion basis compared to the sensitivity estimates on a per-patient basis. Specificity estimates on a per-patient basis were comparable between the four modalities.
Source...