Neuroblastoma is the most prevalent extracranial solid tumor in children, representing 7–8% of all pediatric cancers (1). Approximately 50% of patients are classified as high-risk at the time of diagnosis (2). Despite intensive multimodal treatment, high-risk patients with neuroblastoma experience refractory disease or relapse (3). Peptide receptor radionuclide therapy (PRRT) targets the somatostatin receptor (SSTR) and delivers targeted radiation to SSTR-expressing cells in vivo (4). Among PRRT agents, ¹⁷⁷Lu-DOTATATE has demonstrated significant efficacy and gained regulatory approval in many regions, particularly for SSTR-positive neuroendocrine tumors (NETs) (5, 6). The National Comprehensive Cancer Network (NCCN) (7) has identified PRRT as a therapeutic option for patients with advanced metastatic gastrointestinal, bronchopulmonary, and thymic NETs. Given its success in NETs, there is growing interest in exploring the utility of ¹⁷⁷Lu-DOTATATE and similar SSTR-targeting radiopharmaceuticals in other SSTR-expressing malignancies, including neuroblastoma. Despite of the increasing use of PRRT for neuroblastoma, data on its efficacy are limited. The disease control rate (DCR) varies across studies, and reliable indicators for predicting treatment responses are lacking. Gains et al. (8) evaluated six children with neuroblastoma who underwent two or three cycles of PRRT, yielding a DCR of 83.33%. However, another study observed no objective responses in 20 children who underwent one to four cycles of PRRT (9). Further, Malcolm et al. (10) reported a DCR of 100% (n = 6) for neuroblastoma after four PRRT cycles. These discrepancies may be attributed to the heterogeneity among patient populations, differences in treatment regimens, and assessment criteria. ⁶⁸Ga-DOTATATE positron emission tomography/computed tomography (PET/CT) enables the visualization of tumor SSTR2 expression using radiolabeled somatostatin analogs (11). In comparison to conventional SSTR scintigraphy (e.g., ¹¹¹In-octreotide scanning), ⁶⁸Ga-DOTATATE PET/CT offers superior spatial resolution, higher detection rates for minute lesions, and enhanced quantitative capabilities (12). SSTR PET/CT, represented by ⁶⁸Ga-DOTATATE PET/CT, allows for direct visualization of PRRT targets. Unlike ¹⁸F-FDG PET/CT, it provides precise molecular-level guidance for PRRT, thereby establishing it as a pivotal instrument in the theranostic paradigm. An in-depth analysis of SSTR PET/CT parameters may assist in identifying PRRT-sensitive patients.

Multiple studies have revealed that baseline ⁶⁸Ga-PET standardized uptake values (SUVs), volumetric parameters, and heterogeneity parameters may aid in predicting the response to PRRT in patients with NETs. Previous studies have demonstrated that the maximum SUV (SUV_max) can help predict the treatment response and progression-free survival (PFS) (13–17). However, others found no significant correlation between SUV_max and the treatment response (18, 19). Ortega et al. (14) demonstrated that an elevated ratio of tumor SUV_max to liver SUV (SUV_T/L) on baseline PET serves as a predictor for extended-progression PFS, whereas Durmo et al. (20) identified a correlation between increased tumor volume (TV) and diminished overall survival. Laudicella et al. (21) and Atkinson et al. (22) found that heterogeneity parameters, viz. skewness, kurtosis, and entropy, can help predict treatment response. However, the use of ⁶⁸Ga-labeled tracers is limited by issues such as low production yield, short half-life, and high cost (23, 24). Compared to ⁶⁸Ga-labeled somatostatin analogues, ¹⁸F-AlF-NOTATATE offers significant advantages, including a longer half-life, higher production yield, and superior image resolution, positioning it as a promising replacement with enhanced diagnostic performance (25–27). To our knowledge, no study has investigated the potential of ¹⁸F-AlF-NOTATATE PET/CT in predicting PRRT efficacy in neuroblastoma patients.

Therefore, we aimed to assess the value of SUVs, volumetric parameters, and heterogeneity parameters derived from ¹⁸F-AlF-NOTATATE PET/CT for predicting the efficacy of PRRT in patients with neuroblastoma, to assist in the clinical screening of patients who may benefit from ¹⁷⁷Lu-DOTATATE PRRT.

The NCCN recommends PRRT for treating advanced metastatic NETs (7). Although its application in neuroblastoma has gradually increased in recent years, efficacy data are limited; DCRs vary, and reliable efficacy prediction indicators are lacking. Several studies have shown that the SUV_max derived from ⁶⁸Ga-DOTATATE/TOC PET/CT can predict PRRT response and PFS (13–17). The results of previous studies further support the predictive value of imaging parameters (14, 20, 22). Compared to ⁶⁸Ga-labled somatostatin analogues, ¹⁸F-AlF-NOTATATE offers significant advantages, positioning it as a promising replacement with enhanced diagnostic performance. However, currently, studies on the application of ¹⁸F-AlF-NOTATATE PET/CT in predicting PRRT efficacy in patients with neuroblastoma are lacking. Thus, to our knowledge, ours is the first study to investigate the potential of ¹⁸F-AlF-NOTATATE PET/CT for forecasting the effectiveness of PRRT in patients with neuroblastoma to assist in the clinical assessment of candidates who may be suitable for PRRT.

We chose ¹⁸F-AlF-NOTATATE PET/CT to evaluate the effectiveness of PRRT. Although the NCCN guidelines recommend ¹²³I-MIBG SPECT/CT as the preferred method for assessing metastatic neuroblastomas (32), these two imaging modalities reflect fundamentally different biological mechanisms. ¹²³I-MIBG targets the norepinephrine transporter, whereas ¹⁸F-AlF-NOTATATE targets SSTR (primarily SSTR2). Given that our study aims to evaluate the efficacy of SSTR-targeted PRRT, selecting a PET tracer that also targets SSTR ensures alignment between the diagnostic assessment and the therapeutic target. This allows for a more direct and precise reflection of the tumor’s response to the treatment. Furthermore, compared to ¹²³I-MIBG SPECT/CT, SSTR PET/CT offers superior spatial resolution and higher sensitivity for detecting bone metastases, which further enhances its reliability as a tool for assessing therapeutic efficacy (33). In this study, the DCR of patients treated with one cycle was 20%, and that of patients treated with two cycles was 64.71%, approximating the results of previous studies (8, 9, 34), which reported DCRs of 42.86–100%. Most studies incorporated small sample sizes and retrospective designs, with great inconsistency in clinical questions, inclusion criteria, study design, treatment regimens and survival evaluation. Currently, standardized and effective evaluation criteria for SSTR PET response evaluation are lacking (35, 36).

The reduction in SSTR agonist binding may result from a decrease in SSTRs due to disease progression, treatment effects, or factors like altered perfusion or dedifferentiation (37). Given that the majority of residual lesions were located in the bones and bone marrow, we employed the EORTC criteria for efficacy evaluation. This choice was made because CT/MRI is insufficient for assessing bone marrow lesions, which may not show significant reduction even in the absence of viable tumors (32). Since all patients were children, treatment plans and evaluations required careful consideration. Owing to parental preferences and compassionate care, invasive procedures such as bone marrow aspiration and biopsy could not be performed for all patients (38), complicating the application of INRC criteria (39).

Baseline PET indices such as SUV_max, SUV_mean, and SUV_T/L did not significantly differ between the non-PD and PD groups, differing from previous NET studies (14). However, SUV_T/S and TLA were significantly higher in PD lesions, though their predictive performance was low. The predictive accuracy of SUV_T/S and TLA is limited due to intra-tumor and inter-individual heterogeneity, as well as the complex mechanisms of PRRT. In this study, we observed that the lesion TV in the PD group exceeded that in the non-PD group, which is in agreement with previous results (20, 28). The efficacy of ¹⁷⁷Lu-DOTATATE is influenced by factors such as tumor size due to the limited penetration of the beta emission of ¹⁷⁷Lu (approximately 0.23 mm) (40). Larger tumors, which often exhibit poor blood supply, are more resistant to radiation (41). Therefore, tumor size and other factors must be considered when planning PRRT for optimal outcomes.

In contrast to previous studies (14, 15, 42), we discovered that the non-PD group showed significantly lower SUV_T/L and SUV_max of target lesions on mid-treatment PET compared to the PD group. We postulate that this result may be attributable to tumor cell dedifferentiation or the proliferation of SSTR-negative tumor cells during the progression of neuroblastoma. These tumor cells would exhibit lower tracer uptake on SSTR PET/CT, which reflects a loss of the therapeutic target rather than a true tumor response, even as the overall tumor burden increases. Furthermore, neuroblastoma presents with significant biological differences from NET in terms of origin, epidemiology, sites of involvement, and clinical presentation. The potential for selection bias, given that our patient cohort had predominant bone and bone marrow involvement, coupled with insufficient statistical power due to a small sample size, may also be contributing factors to this contrary finding. Laudicella et al. (21) conducted a bone area subgroup analysis in patients with NETs, revealing that SUV_max in responders was significantly lower than that in non-responders, consistent with our findings. Moreover, the mid-treatment SUV_mean and the mid-to-baseline SUV_mean ratio in this study demonstrated better predictive accuracy for treatment efficacy, corroborating with previous results (29, 43). However, Durmo et al. (20) and Werner et al. (44) reported that SUV_mean could not reliably predict treatment response or survival in patients undergoing PRRT. SUV_mean provides a more comprehensive reflection of tumor lesions than SUV_max but is highly influenced by inter-observer variability in tumor delineation. Semi-automatic delineation can mitigate this issue, but SUV_mean should not be utilized as the sole parameter for patient screening in PRRT; rather, it can serve as a valuable prognostic factor. This highlights that moving toward normalized or corrected parameters may enhance predictive power. Ilan et al. (45) noted that corrected SUV_max values, including tumor-to-blood, tumor-to-spleen, and tumor-to-liver ratios, are more dependable measures than the absolute SUV_max.

The interplay between cancer cells and the microenvironment during tumorigenesis can greatly influence tumor invasiveness and resistance to treatment (46). Tumors with greater internal heterogeneity often have a worse prognosis. However, no notable differences in heterogeneity parameters were observed between responders and non-responders at baseline or mid-treatment. A key limitation that likely contributes to this finding is that our analysis was confined to first-order, histogram-based texture features (coefficient of variation, skewness, and kurtosis). These metrics describe the statistical distribution of voxel intensities within a tumor but do not capture the spatial relationships or arrangement of those voxels. This approach omits higher-order textural features, such as entropy and those derived from the Gray-Level Co-occurrence Matrix like homogeneity and contrast. Such metrics provide a more sophisticated characterization of textural patterns and have been successfully used in other studies to predict PRRT response (14, 21, 22, 44). It is plausible that while the overall intensity distributions were similar between groups in our cohort, underlying differences in spatial heterogeneity could exist, which our first-order analysis was not designed to detect. Therefore, the lack of significant findings for heterogeneity should be interpreted with caution. Future research incorporating a more comprehensive panel of higher-order radiomic features is warranted to fully explore the predictive value of intratumoral heterogeneity in this setting.

The results of this study demonstrate that the ratios of SUV_max, SUV_mean, SUV_T/L, and TLA between interim and baseline PET scans exhibit favorable predictive performance, with the SUV_max ratio showing superior performance among all parameters. These ratios reflect dynamic changes in tumor surface receptor expression before and after treatment, suggesting that early-to-interim PET imaging may serve as a complementary tool for therapeutic efficacy prediction. Furthermore, our findings indicate that dynamic metrics (e.g., multi-timepoint variations) could provide more accurate reflection of tumor response compared to static single-timepoint measurements. However, it is noteworthy that comparative analysis between baseline and interim PET revealed significant increases in single-lesion SUV_max at interim assessment in seven patients. This observation should be interpreted with caution. A potential confounding factor is the initial use of ¹⁷⁷Lu-DOTATOC therapy with relatively lower SSTR2 affinity in all seven cases (47). Despite such influencing factor, it does not diminish our central conclusion. Rather, it reinforces the idea: implementing multi-timepoint PET analysis to evaluate dynamic ratios can enhance the precision of response assessment and provide a stronger basis for personalized treatment.

This study has some limitations that must be considered when interpreting the findings. First, the primary constraint is the study’s single-center design and small sample size, which diminishes the statistical power of our findings and may limit their generalizability. Second, this limitation is compounded by significant heterogeneity within the study cohort. Specifically, while most patients received ¹⁷⁷Lu-DOTATATE, some were treated with ¹⁷⁷Lu-DOTATOC, an agent with a relatively lower SSTR2 affinity. As noted in our discussion, this represents a critical confounding factor. In addition, our cohort was predominantly composed of patients with bone and bone marrow metastases, with insufficient representation of soft-tissue lesions. Consequently, the predictive models and conclusions drawn from this study may be most applicable to bone-dominant disease, and their relevance for neuroblastoma patients with primarily soft-tissue involvement remains uncertain. Third, the study lacks a histopathological “gold standard” for validation. Due to ethical and practical challenges associated with performing invasive biopsies in a pediatric population, our efficacy evaluation relied mainly on imaging-based criteria. Therefore, we cannot definitively confirm whether changes in SUV values correspond directly to a change in viable tumor cell count or were influenced by factors such as treatment-related inflammation. Fourth, the limited number of treatment cycles is a constraint, as many patients were unable to complete four treatment cycles due to terminal illness or other barriers. Furthermore, our textural analysis was limited to first-order features, potentially obscuring more complex patterns of spatial heterogeneity that higher-order metrics could have revealed. Lastly, as originally stated, the limited number of treatment cycles and the short follow-up period precluded correlation analyses with long-term clinical outcomes like PFS and overall survival. Future multi-center, prospective studies with standardized treatment and imaging protocols, coupled with long-term follow-up data, are essential to validate and refine our preliminary findings.

Quantitative parameters based on ¹⁸F-AlF-NOTATATE PET/CT have great potential in predicting PRRT response in pediatric neuroblastoma. An interim-to-baseline PET lesion SUV_max ratio of ≤1.25 can effectively predict the response of that lesion to PRRT, providing a reliable basis for evaluating PRRT efficacy and implementing personalized treatment.