Comparing the diagnostic efficacy of [¹⁸F]FDG PET/CT and [¹⁸F]FDG PET/MRI for detecting bone metastases in breast cancer: a meta-analysis

An extensive search was conducted in the PubMed, Embase, Web of Science, and Cochrane Library databases to identify available publications up to February 2023. The search was conducted using the following keyword terms: “Positron-Emission Tomography”, “Breast Neoplasms” and “Bone metastases”. More details could be found in the Supplementary Table 1. The reference lists of the included studies were manually searched to find additional relevant articles.

Studies were included in this meta-analysis if they evaluated the diagnostic performance of [¹⁸F]FDG PET/CT and/or [¹⁸F]FDG PET/MRI in patients with breast cancer bone metastases with a sample size of more than 10 patients.

Duplicated articles, abstracts without full texts, editorial comments, letters, case reports, reviews, meta-analyses, irrelevant titles and abstracts, and non-English full-text articles were excluded. Studies with incomplete or unclear data necessary to calculate the sensitivity or specificity of the imaging modality being studied were excluded. In addition, studies using PET without CT or MRI, or using different radiotracers were excluded. For studies using the same data, only the latest studies were taken into consideration.

Two researchers independently read the titles and abstracts of retrieved articles using the predetermined selection criteria. Subsequently, full-text evaluation was conducted to ascertain each study's eligibility. The event of discrepancies between the researchers were resolved through discussion, ultimately arriving at a consensus.

Two researchers independently assessed the quality of the included studies utilizing the Quality Assessment of Diagnostic Performance Studies (QUADAS-2) tool.¹³ The QUADAS-2 tool encompasses four essential domains: (1) patient selection; (2) index test; (3) reference standard; and (4) flow and timing. The risk of bias was rated as “high risk,” “low risk,” or “unclear risk.”

In assessing the risk of bias, several key aspects were evaluated. First, patient selection bias was addressed by enrolling consecutive patients. Second, the results of the index test were evaluated independently of the outcomes of the reference standard to minimize potential bias. Third, the reference standard was evaluated without knowledge of the results of the index test to ensure objectivity. Finally, the flow and timing aspect examined the appropriateness interval (less than 3 months) between the index tests and the reference standard. Regarding applicability concerns, the analysis focused on three main questions. First, patient selection was “Are there any concerns regarding the relevance of the included patients to the scope of the review?” Second, the index test was “Are there concerns that the target condition as defined by the reference?” Third, the reference standard was “Are there concerns about the compatibility between the target condition, as established by the reference standard, and the review question?”

Two researchers independently extracted data from all the included articles. The gathered data included information about the author, year of publication, and the type of imaging test used in the study, study features (country, study design, analysis, and reference standard), characteristics of patients (number of patients, clinical indication, mean/median age, and previous treatment), and technical aspects (scanner modality, ligand dose, and image analysis).

In cases of disagreements, the researchers discussed the issue until a consensus was reached to ensure accuracy in the extracted data.

The main outcome measure were the sensitivities and specificities of [¹⁸F]FDG PET/CT and [¹⁸F]FDG PET/MRI in patient-based analysis, lesion-based analysis and head-to-head comparison. Sensitivity was defined as the ratio of patients or lesions with true positive (TP) scans to the sum of TP and false negative (FN) scans for either patients or lesions have been reported; Specificity was defined as the ratio of patients or lesions with true negative (TN) scans to the sum of TN scans and false negative (FN) scans have been reported.

Sensitivity and specificity were assessed using the DerSimonian and Laird method, followed by transformation via the Freeman-Tukey double inverse sine transformation. The Jackson method was used to calculate the confidence intervals. The Cochrane Q and I² statistics were used to assess the heterogeneity within and between groups.¹⁴ If the heterogeneity between the studies differed significantly (P < 0.10 or I² > 50%), sensitivity analysis was performed by reassessing the sensitivities or specificities following the omission of articles one by one. This was done to evaluate the robustness of the overall sensitivities or specificities and to identify single studies that may contribute to heterogeneity.

We evaluated publication bias by employing both funnel plot and Egger's test (for outcomes including over 10 studies). For all statistical tests except heterogeneity (P < 0.10), a significance level of P < 0.05 was considered statistically significant. Statistical analyses were conducted using R software version 4.1.2 for statistical computing and graphics.

The preliminary search revealed a total of 1525 publications. However, 542 studies were considered duplicates, and another 950 did not meet the eligibility criteria and were therefore not included in the study. After a comprehensive review of the full texts of the remaining 33 articles, another 17 were deemed ineligible for the study either because data (TP, FP, FN, and TN) were not available (n = 8) or the radiotracer was different (n = 3). In addition, non-English articles (n = 2) and PET without CT or MRI articles (n = 3) were excluded. Finally, 16 articles^{9,10,11,15,16,17,18,19,20,21,22,23,24,25,26,27} (including 4 head-to-head comparison articles) evaluating the diagnostic efficacy of [¹⁸F]FDG PET/CT (n = 15)^{17,18,19,20,22,23,24,25,26,27}, and [¹⁸F] FDG PET/MRI (n = 5)^{9,10,11,15,16} were included in the meta-analysis. The article selection process, according to the PRISMA flow diagram, is depicted in Figure 1.

FIGURE 1.

The 16 eligible studies included a total of 1,261 breast cancer patients (range from 18 to 225). Among the included studies, 9 articles were retrospective studies, while 7 articles were prospective studies. In terms of analysis methods, 3 articles employed patient-based analysis, 9 articles used lesion-based analysis, and 4 articles utilized both methods. 2 articles used pathology as the reference standard, 11 articles employed pathology and/or follow-up imaging as the reference standard, and 3 articles solely relied on follow-up imaging as the reference standard. Regarding clinical indications, 3 articles involved patients exclusively at the initial stage, 6 articles included patients only at the post-treatment stage, and the remaining 7 articles included patients at both initial and post-treatment stages. Table 1 and Table 2 summarize the study and patient characteristics of [¹⁸F]FDG PET/CT and [¹⁸F]FDG PET/MRI, while Supplementary Table 2 and Supplementary Table 3 present the technical aspects.

The risk of bias for each study according to the QUADAS-2 tool is illustrated in Figure 2. For the patient selection risk of bias assessment, we found 2 studies that graded as “high risk” since they didn’t include consecutive patients. For the index test, 3 studies were graded as “high risk” since the applied cut-off values were not pre-determined. With regards to the reference standard, 2 studies were graded as “high risk” as the final diagnosis was not determined independently by two or more physicians. The flow and timing standard were graded as “high risk” in 2 studies because some participants were excluded from data analyses. There were no major concerns with the quality of the included studies based on the overall quality assessment.

FIGURE 2.

For patient-based analysis, a total of 8 studies with 213 patients were included in the analysis, and the pooled sensitivity of [¹⁸F]FDG PET/CT in detecting bone metastases in breast cancer was 0.73 (95% CI: 0.42–0.96), whereas [¹⁸F]FDG PET/MRI had an overall sensitivity of 0.99 (95% CI:0.90–1.00) (Figure 3). There was significant difference between [¹⁸F]FDG PET/CT and [¹⁸F]FDG PET/MRI in the sensitivity (P = 0.04) (Figure 3). After removing Hahn et al.'s study¹⁸ in our sensitivity analysis, the I² value became 0%, suggesting it may be the potential source of heterogeneity. However, the results from the sensitivity analysis remained stable, with only minor variations observed, ranging from 0.66 to 0.86 (Supplementary Figure 1).

FIGURE 3.

For lesion-based analysis, a total of 13 studies with 1588 lesions were included in the analysis, and the pooled sensitivity of [¹⁸F]FDG PET/CT in detecting bone metastases in breast cancer was 0.89 (95% CI: 0.80–0.96), whereas [¹⁸F]FDG PET/MRI had an overall sensitivity of 0.99 (95% CI:0.96–1.00) (Figure 4). There was significant difference between [¹⁸F]FDG PET/CT and [¹⁸F]FDG PET/MRI in the overall sensitivity (P < 0.01) (Figure 4). Regarding the pooled overall sensitivity of [¹⁸F] FDG PET/CT in lesion-based analysis, the I² was 94%. The sensitivity analysis revealed no potential source of heterogeneity. The results following sensitivity analysis remained stable, and only minor variations in the results ranging from 0.88 to 0.92 were noted (Supplementary Figure 2). The funnel plot and Egger's test revealed no evidence of publication bias for [¹⁸F]FDG PET/CT in lesion-based analysis (P = 0.30) (Supplementary Figure 3).

FIGURE 4.

For head-to-head comparison, a total of 4 studies with 442 patients or lesions were included in the analysis, and the pooled sensitivity of [¹⁸F]FDG PET/CT in detecting bone metastases in breast cancer was 0.87 (95% CI: 0.77–0.94), whereas [¹⁸F]FDG PET/MRI had an overall sensitivity of 0.99 (95% CI:0.96–1.00) (Supplementary Figure 4). A significant difference was observed in the overall sensitivity between [¹⁸F]FDG PET/CT and [¹⁸F] FDG PET/MRI. (P < 0.01) (Supplementary Figure 4). Regarding the pooled overall sensitivity of [¹⁸F]FDG PET/CT in lesion-based analysis, the I² was 64%. After removing Botsikas et al.'s study⁹ in our sensitivity analysis, the I² value became 0%, suggesting it may be the potential source of heterogeneity. However, the results from the sensitivity analysis remained stable, with only minor variations observed, ranging from 0.83 to 0.90. (Supplementary Figure 5).

For patient-based analysis, a total of 7 studies with 625 patients were included in the analysis, and the pooled specificity of [¹⁸F]FDG PET/CT in detecting bone metastases in breast cancer was 1.00 (95% CI: 0.97–1.00), whereas [¹⁸F]FDG PET/MRI had an overall specificity of 1.00 (95% CI:0.98–1.00) (Figure 5). There was no significant difference between [¹⁸F]FDG PET/CT and [¹⁸F]FDG PET/MRI in the overall specificity (P = 0.55) (Figure 5). The pooled overall specificity of [¹⁸F]FDG PET/CT and PET/MRI exhibited I² values of 52% and 54%, respectively. Sensitivity analysis revealed that removing Niikura et al.'s study²¹ reduced PET/CT heterogeneity (I² = 0%), while removing Botsikas et al.'s study⁹ had a similar effect on PET/MRI. Nonetheless, both analyses yielded stable results, with minor variations between 0.99 and 1.00 (Supplementary Figures 6 and 7).

FIGURE 5.

For lesion-based analysis, a total of 9 studies with 1023 lesions were included in the analysis, and the pooled specificity of [¹⁸F]FDG PET/CT in detecting bone metastases in breast cancer was 0.99 (95% CI: 0.97–1.00), whereas [¹⁸F]FDG PET/MRI had an overall specificity of 0.99 (95% CI:0.95–1.00) (Figure 6). A significant difference was observed in the overall specificity between [¹⁸F]FDG PET/CT and [¹⁸F]FDG PET/MRI (P = 0.07) (Figure 6). Regarding the pooled overall specificity of [¹⁸F]FDG PET/CT in lesion–based analysis, the I² was 67%. After removing Hahn et al.'s study¹⁸ in our sensitivity analysis, the I² value became 49%, suggesting it may be the potential source of heterogeneity. However, the results from the sensitivity analysis remained stable, with only minor variations observed, ranging from 0.99 to 1.00. (Supplementary Figure 8).

FIGURE 6.

For head-to-head comparison, a total of 2 studies with 466 patients or lesions were included in the analysis, and the pooled specificity of [¹⁸F]FDG PET/CT in detecting bone metastases in breast cancer was 1.00 (95% CI: 0.98–1.00), whereas [¹⁸F]FDG PET/MRI had an overall specificity of 0.98 (95% CI:0.91–1.00) (Supplementary Figure 9). No significant difference was observed in the overall specificity between [¹⁸F]FDG PET/CT and [¹⁸F]FDG PET/MRI (P = 0.50) (Supplementary Figure 9).

In the 4 head-to-head comparison studies, one study (Melsaether et al.) did not provide information regarding the complementary role of PET/CT and PET/MRI in identifying bone metastases.¹⁰ Therefore, the evaluation was from 3 studies (Supplementary Table 4). Among these studies, PET/MRI correctly identified bone metastases in 10 out of 98 patients or lesions (10.2%) with initially negative PET/CT results. Conversely, PET/CT correctly identified bone metastases in none of the 86 patients (0%) with initially negative PET/MRI results.

Furthermore, 2 studies reported information on the detection of bone metastases in breast cancer patients using MRI_(PET/MRI) alone, while 4 studies provided data on CT_(PET/CT) alone. The results indicate that MRI_(PET/MRI) alone demonstrated a higher detection rate (65.5%, 180 out of 275) compared to CT_(PET/CT) alone (51.2%, 166 out of 324) (Supplementary Table 4).