PubMed, Embase, Web of Science, and Cochrane Library were searched from January 1985 to June 2023 for relevant studies, with language restriction to English. Search items mainly contained the following: (“Neuroblastoma” or “Neuroblastomas”) and (“Positron Emission Tomography Computed Tomography”). The detailed search strategy is shown in the Supplementary Material.

Studies meeting the following criteria were included: observational study (prospective and retrospective) or clinical trial that applied ¹⁸F-FDG PET/CT and relevant parameters (SUVmax, MTV, and TLG) in NB patients and reported survival data, such as OS, progression-free survival (PFS), and event-free survival (EFS).

Literature review, conference summary, case report, and editorial materials were excluded.

Literature search and study selection were conducted by two reviewers independently, and disagreements were settled via discussion.

Quality assessment of included studies was performed by two reviewers independently using the Quality in Prognostic Studies (QUIPS) tool (16) via Review Manager 5.4 software. QUIPS contains six domains: study participation, study attrition, measurement of prognostic factors, measurement of outcome, study confounding, and statistical analysis and reporting. Disagreements were settled via discussion.

Data were extracted independently by two reviewers using a pre-designed form that included the following: name of the first author, publication date, sample size, country, study design, characteristics of participants (gender distribution, tumor grade, tumor site, treatment after PET/CT scans, volumes of interest (VOIs) for recording SUVmax, and reported survival), PET parameters, and cutoff values of parameters.

The primary endpoint was OS, defined as the time interval from the initiation of treatment to all-cause death. The secondary outcome was PFS, referring to recurrence-free survival and the time interval from the date of treatment initiation to tumoral recurrence or metastasis. EFS was calculated from diagnosis to the first occurrence of relapse, progression, secondary malignancy, death, or the last contact if no event occurred. Hazard ratio (HR) was applied as the statistic for the association of SUVmax, MTV, or TLG with PFS, EFS, and OS. PFS, EFS, or OS data were extracted using methods mentioned previously (17). Univariate or multivariate HR with a 95% confidence interval (95% CI) were extracted from each study if provided; otherwise, Engauge Digitizer would be applied (http://markummitchell.github.io/engauge-digitizer/) to estimate the survival rate through Kaplan–Meier curve and reconstruct HR estimate and its variance, assuming that patients were censored at a constant rate during the follow-up. A heterogeneity test was performed using chi-square (χ²) test and I²statistic (18). I² less than 50% with a p-value not less than 0.1 indicated no significant heterogeneity among the studies, and a fixed-effects model would be applied; otherwise (I² greater than 50% with a p-value less than 0.1), a random-effects model would be applied. Meanwhile, sensitivity analysis was performed by removing each included study one by one to assess the robustness of the results. Statistical analysis was performed using Stata Version 16.0 (College Station, TX, USA). A p-value less than 0.05 indicated statistical significance.

The flow diagram of the study selection process is presented in Figure 1. A total of eight studies, involving 325 participants, were included, among which seven studies (13, 19–24) were retrospective design and one study (25) was prospective. According to the INRGSS grade, one study (25) only recruited patients with stage IV neuroblastoma; one study recruited those at stages I, II, and IV (20); four studies recruited patients at all grades (13, 19, 21, 24); the remaining two studies failed to clearly describe the grading of the patients (22, 23). There were six studies that included neuroblastoma originating in the adrenal glands, retroperitoneum, and mediastinum (19–24), and the other two studies (13, 25) failed to clearly state the tumor sites. The characteristics of included studies are shown in Table 1. All the studies used ¹⁸F-FDG for PET imaging, among which seven studies reported SUVmax (13, 19, 21–25), four studies reported MTV and TLG (13, 20, 22, 24), four studies reported the predictive value of SUVmax for OS (19, 21, 23, 25), three studies reported association of SUVmax with PFS (or recurrence-free survival) (13, 19, 22), two studies reported association of SUVmax with EFS (21, 24), two studies reported association of MTV and TLG with OS (20, 22), and three studies reported the predictive value of MTV and TLG for PFS (13, 20, 22). One study (19) provided spheroid-shaped VOI for the primary tumor lesion and metastatic lesions of each patient to evaluate FDG uptake of neuroblastoma lesions, and SUVmax in each VOI was measured, while another study measured SUVmax in VOI for the most intense lesion (25). The cutoff value of SUVmax ranged from 3.31 to 12.01, and those of MTV in two studies (20, 22) were 88.1 and 191 cm³, respectively. The cutoff values of TLG in two studies (2, 5) were 1,045.2 and 341.41 g, respectively.

There were four studies (19, 21–23) that were graded as unclear in selection bias due to no description of consecutive selection for participants, one study (13) was graded as high selection bias due to limited sample size, and one study (25) was graded as high selection due to recruitment of only stage IV patients. All the studies were graded as low risk of attrition bias. There was one study graded as unclear risk of bias in prognostic factor measurement (23) because it failed to state the participation of two experienced nuclear medicine physicians in the measurement. There were five studies (13, 19–22) graded as having unclear risk of bias in outcome measurement due to no description of detailed methods for measurement. There were four studies (13, 19, 24, 25) graded as high risk in confounding bias due to the lack of multivariate analysis and one study (22) due to an unclear risk because it performed both multivariate analysis and univariate survival analysis. In terms of statistical analysis and reporting, six studies (13, 19–22, 25) were graded as high risk of bias in that these studies failed to provide the HRs of non-significant factors. The overall quality of included studies was moderate (Figure 2).

There were four studies that reported an association of SUVmax with OS (19, 21, 23, 25). No significant heterogeneity was considered among the studies (I^2 = 1.5%), followed by a fixed-effects model applied. Meta-analysis showed that the value of SUVmax was negatively associated with the OS of NB patients [HR = 1.27, 95% CI (1.11, 1.45), p = 0.001] (Figure 3A).

There were three studies that reported the association of SUVmax with PFS (13, 19, 22). No significant heterogeneity was considered among the studies (I^2 = 46.8%, p = 0.153), followed by the fixed-effects model applied. Meta-analysis showed no significant association between SUVmax and the PFS of NB patients [HR = 1.03, 95% CI (0.99, 1.07), p = 0.222] (Figure 3B).

There were two studies that reported the association of MTV with OS (20, 22). Significant heterogeneity was observed (I² = 88.1%, p = 0.004), so a random-effects model was used. Meta-analysis showed no significant association between MTV and the OS of NB patients [HR = 2.46, 95% CI (0.34, 18.06), p = 0.376] (Figure 3C).

There were three studies that reported an association of MTV with PFS (13, 20, 22). Significant heterogeneity was observed (I² = 85.8%, p = 0.001), and a random-effects model was used. Meta-analysis showed no significant association between MTV and the PFS of NB patients [HR = 2.60, 95% CI (0.68, 9.88), p = 0.161] (Figure 3D).

There were two studies that reported the association of TLG with OS (20, 22). No significant heterogeneity was considered among the studies (I² = 43.4%, p = 0.184), and a fixed-effects model was applied. Meta-analysis showed that TLG was significantly associated with the OS of NB patients [HR = 0.99, 95% CI (0.99, 0.99), p = 0.00] (Figure 3E).

There were three studies that reported the association of TLG with PFS (13, 20, 22). Significant heterogeneity was observed (I² = 87.6%, p = 0.000), and a random-effects model was applied. Meta-analysis showed no significant association between TLG and the PFS of NB patients [HR = 1.00, 95% CI (1.00, 1.00), p = 0.974] (Figure 3F).

There were two studies that reported the association of SUVmax with EFS (21, 24). Significant heterogeneity was observed (I² = 86.2%, p = 0.007), so a random-effects model was used. Meta-analysis showed no significant association between SUVmax and the EFS of NB patients [HR = 2.58, 95% CI (0.37, 18.24), p = 0.341] (Figure 3G).

A sensitivity analysis was performed (Supplementary Figure S1) to assess the robustness of the results. Since the research data on OS based on MTV, EFS based on SUVmax, and OS based on TLG are relatively small, only sensitivity analysis was performed on OS based on SUVmax, PFS based on SUVmax, PFS based on MTV, and PFS based on TLG. Among studies of SUVmax on OS, the combined HRs were found to be stable, suggesting that no individual study significantly affected the results (Supplementary Figure S1A). Of all studies of SUVmax on PFS, one study (22) had a great impact on the results. After this study was excluded, the combined HR was far larger than before. As for studies of MTV on PFS, the combined HRs were also found to be stable, indicating that no individual study significantly affected the results. Among studies of TLG on PFS, after excluding one study (13), the value of HR remained unchanged. This indicated that this study had no effect on the results.