Detecting Bone Marrow Involvement in Hodgkin Lymphoma
Detecting Bone Marrow Involvement in Hodgkin Lymphoma
The computer-aided search revealed 230 articles from PubMed/Medline and 235 articles from Embase ( Table 1 ). After discarding duplicates, 419 articles remained and were screened on title and abstract for eligibility in this meta-analysis. Reviewing titles and abstracts from Embase revealed 49 articles potentially eligible for inclusion. After reviewing the full article, 23 articles were excluded because stand-alone FDG-PET was carried out rather than integrated FDG-PET/CT, 11 articles were excluded because they provided insufficient data to calculate the proportion of FDG-PET/CT-negative cases with a positive BMB among all cases, two studies were excluded because they enrolled <10 patients with Hodgkin lymphoma, two studies were excluded because of mixing newly diagnosed with previously treated Hodgkin lymphoma, one study was excluded because it only enrolled patients who were positive for bone marrow involvement at FDG-PET/CT, and one study was excluded because the same data were used in another article providing more study details. Thus, nine studies remained, comprising a total sample size of 955 patients with newly diagnosed Hodgkin lymphoma. Seven of these nine studies used at least both BMB and follow-up FDG-PET/CT studies as reference standard, whereas two of these nine studies only used BMB as reference standard. Characteristics of included studies are shown in Table 2 and Table 3 .
The QUADAS-2 scores are displayed in Table 4 . The risk of bias for the index test and reference standard remained unclear in the majority of cases since these studies did not report whether the index test (six of nine studies) and/or reference standard (all studies) were interpreted in a blinded manner, and two of nine studies did not report the applied criteria for FDG-PET/CT positivity. There were no major concerns regarding the applicability of included studies.
Seven of nine included studies allowed calculation of sensitivity and specificity of FDG-PET/CT. The sensitivity and specificity of FDG-PET/CT for the detection of bone marrow involvement ranged from 87.5% to 100% and from 86.7% to 100%, respectively ( Table 5 ). Spearman ρ between the logit of sensitivity and the logit of 1 − specificity was −0.679 (P = 0.094), which suggested that there was no threshold effect. Pooled sensitivity and specificity of FDG-PET/CT were 96.9% [95% CI 93.0% to 99.0%] and 99.7% (95% CI 98.9% to 100%), respectively ( Table 5 ). The area under the sROC curve was 0.9860 (Figure 1). The DORs were homogeneous across individual studies (I = 32.8%). Therefore, no further subgroup analyses were carried out. The proportions of FDG-PET/CT-negative cases with a positive BMB among all cases were homogeneous across all nine included studies (I = 34.3%). The weighted summary proportion (fixed effects model) of FDG-PET/CT-negative patients with a positive BMB among all cases was 1.1% (95% CI 0.6% to 2.0%).
(Enlarge Image)
Figure 1.
ROC curves show sensitivities and specificities of individual studies in ROC space with sROC and 95% confidence interval.
Results
Literature Search
The computer-aided search revealed 230 articles from PubMed/Medline and 235 articles from Embase ( Table 1 ). After discarding duplicates, 419 articles remained and were screened on title and abstract for eligibility in this meta-analysis. Reviewing titles and abstracts from Embase revealed 49 articles potentially eligible for inclusion. After reviewing the full article, 23 articles were excluded because stand-alone FDG-PET was carried out rather than integrated FDG-PET/CT, 11 articles were excluded because they provided insufficient data to calculate the proportion of FDG-PET/CT-negative cases with a positive BMB among all cases, two studies were excluded because they enrolled <10 patients with Hodgkin lymphoma, two studies were excluded because of mixing newly diagnosed with previously treated Hodgkin lymphoma, one study was excluded because it only enrolled patients who were positive for bone marrow involvement at FDG-PET/CT, and one study was excluded because the same data were used in another article providing more study details. Thus, nine studies remained, comprising a total sample size of 955 patients with newly diagnosed Hodgkin lymphoma. Seven of these nine studies used at least both BMB and follow-up FDG-PET/CT studies as reference standard, whereas two of these nine studies only used BMB as reference standard. Characteristics of included studies are shown in Table 2 and Table 3 .
Methodological Quality Assessment
The QUADAS-2 scores are displayed in Table 4 . The risk of bias for the index test and reference standard remained unclear in the majority of cases since these studies did not report whether the index test (six of nine studies) and/or reference standard (all studies) were interpreted in a blinded manner, and two of nine studies did not report the applied criteria for FDG-PET/CT positivity. There were no major concerns regarding the applicability of included studies.
Diagnostic Performance
Seven of nine included studies allowed calculation of sensitivity and specificity of FDG-PET/CT. The sensitivity and specificity of FDG-PET/CT for the detection of bone marrow involvement ranged from 87.5% to 100% and from 86.7% to 100%, respectively ( Table 5 ). Spearman ρ between the logit of sensitivity and the logit of 1 − specificity was −0.679 (P = 0.094), which suggested that there was no threshold effect. Pooled sensitivity and specificity of FDG-PET/CT were 96.9% [95% CI 93.0% to 99.0%] and 99.7% (95% CI 98.9% to 100%), respectively ( Table 5 ). The area under the sROC curve was 0.9860 (Figure 1). The DORs were homogeneous across individual studies (I = 32.8%). Therefore, no further subgroup analyses were carried out. The proportions of FDG-PET/CT-negative cases with a positive BMB among all cases were homogeneous across all nine included studies (I = 34.3%). The weighted summary proportion (fixed effects model) of FDG-PET/CT-negative patients with a positive BMB among all cases was 1.1% (95% CI 0.6% to 2.0%).
(Enlarge Image)
Figure 1.
ROC curves show sensitivities and specificities of individual studies in ROC space with sROC and 95% confidence interval.
