DISCUSSION
18F-FDG-PET/CT has been formally incorporated into the standard initial staging, interim response evaluation, and residual disease evaluation at the end of treatment for both adult and pediatric patients with HL23-25. PET/CT-based, response-adapted treatment has been proven to improve survival and reduce the risk of long-term toxicity in pediatric HL patients5. It is critical that the response assessment is reliable and reproducible, but the interpretation of interim and post-treatment PET/CT scans can be difficult due to false-positive results8,26, 27. The IHPC and DC are the main criteria for the clinical evaluation of both interim and post-treatment PET/CT13-16. In the present study, we used the IHPC and DC to analyze a homogeneous cohort of pediatric and adolescent HL patients who underwent regular interim PET/CT after 2 cycles of chemotherapy and underwent post-treatment PET/CT. Our study found that post-treatment PET/CT showed a better diagnostic performance than interim PET/CT in the prediction of the outcomes of pediatric and adolescent HL patients. Moreover, the DC-4 cutoff could further improve diagnostic accuracy when compared with the IHPC and DC-3 (Table2).
In our patient cohort, 89 patients underwent interim PET/CT after 2 cycles of chemotherapy, and their PET/CT scan results were analyzed. Given that the index reflecting reduced metabolism is expected to be more discriminating after 2 cycles of chemotherapy than after 4 or 6 cycles28, we excluded the results of irregular PET/CT scans. Our study indicated that interim PET/CT had a relatively low PPV (15%, 19%, and 15% when using the IHPC, DC-3, and DC-4, respectively), but a high NPV (93%, 95%, and 88%, respectively). A retrospective study of 34 children with HL indicated that interim PET/CT had a good PPV of 75% and a high NPV of 96%29. Similar results have been reported for adult patients with HL and NHL, with PPVs of 57%–85% and NPVs of 80%–100%9,30, 31. However, some studies concerning the interim PET/CT visual assessment in pediatric HL patients have showed poor and variable PPVs (ranging between 11% and 30%) and high NPVs (ranging between 80% and 100%)32. The poor PPVs observed in the present study and in previous studies might be explained by high false-positive rates. Like adult lymphomas, cervical lymph node hyperplasia and inflammation show high FDG uptake on PET/CT scans, but are benign processes and cause false-positive results in pediatric and adolescent patients with lymphoma33. The residual 18F-FDG uptake observed on interim PET/CT scans might be related to inflammatory cells. Although the diagnostic accuracy was improved by using DC-4 (75%) rather than IHPC (39%) or DC-3 (55%) in the present study, the diagnostic performance of interim PET/CT was still poor. Moreover, no significant differences in PFS were observed between patients with positive and negative interim PET/CT scans according to the IHPC, DC-3, and DC-4. It should be noted however that both the present study and previous studies concerning interim PET/CT in pediatric and adolescent HL were retrospective in nature and limited by small sample sizes.
Given the results of interim PET/CT in our study, interim PET/CT could not circumvent the need for post-treatment PET/CT evaluation. Thus, we further analyzed the 147 post-treatment PET/CT scans. Improved PPVs were observed using the post-treatment PET/CT scans as compared with the interim PET/CT scans (IHPC: 23% vs. 15%, DC-3: 27% vs. 19%, and DC-4: 33% vs. 15%), indicating that the use of post-treatment PET/CT might reduce false-positive rates. Furthermore, the AUC and accuracy of post-treatment PET/CT were both higher than those of interim PET/CT (Tables 2). Consistent with our findings, a prospective study of 57 pediatric HL patients reported that the predictive value of interim PET/CT was low, and that post-treatment PET/CT evaluated using the DC had better specificity (96% vs. 76%) and PPV (33% vs. 8%) than conventional imaging34. The above finding is probably a consequence of response-adapted treatment in our study: Patients with DC scores of 4 and 5 in the interim PET/CT evaluation received chemotherapy-dose escalation and radiotherapy; hence, the post-treatment PET/CT was more valuable in predicting progression.
In our study, among the three models of PET/CT evaluation, DC-4 showed the highest specificity, AUC, and accuracy, indicating that DC-4 might have a superior diagnostic performance than IHPC and DC-3. Furthermore, Cox regression and Kaplan-Meier analyses revealed that a positive post-treatment PET/CT according to DC-4 was strongly associated with a high risk of progression (hazard ratio: 7.82). A similar conclusion was drawn in a retrospective study of 101 adult patients with lymphoma: the specificity, accuracy, and hazard ratio of positive post-treatment PET/CT were all higher when using the DC-4 rather than the DC-3 and IHPC (specificity: 87% vs.76% vs. 67%, accuracy: 86% vs. 84% vs. 76%, and hazard ratio: 3.2 vs. 0.7 vs. 1.57, respectively)9. However, another retrospective study of 72 pediatric HL patients reported that the IHPC showed a higher specificity (95% vs. 70%), accuracy (89% vs. 46%), and PPV (40% vs. 25%) than the DC in the evaluation of post-treatment PET/CT11. The application of the IHPC depends on the evaluator’s experience and is limited by the quality of image reconstruction35, 36. In contrast, concordance among observers when using the DC has been demonstrated, and discordant interpretations among reviewers occur in only a few challenging cases37. Furthermore, the threshold selected for PET/CT positivity may vary depending on the histological subtype and the treatment. Thus, on the basis of our results, we tentatively regard that the DC-4 had a superior diagnostic performance than the IHPC and DC-3 in the evaluation of post-treatment PET/CT among pediatric and adolescent patients with HL.
Our study has several limitations. First, there was variability in image and data acquisition due to the retrospective nature of our study. Second, some patients received interim PET/CT-based, response-adapted treatment, but the definition of the threshold of positivity varied among doctors. Treatment bias might potentially influence survival. Furthermore, we excluded patients with primary refractory HL, with a consequent impact on overall survival. Further large prospective studies are required to define the best strategy for the use of the DC in specific clinical situations.