Our study provides first and preliminary evidence that not only does quantitative DWI reliably differentiate between malignant and benign neuroblastic tumours, but it may also provide prognostic biomarkers for evaluating therapy response as well as predicting risk of tumour recurrence.
One of the earliest published clinical applications of extracranial DWI in paediatric patients, as reported by Uhl et al. [13] suggested its diagnostic utility in the work-up of neuroblastic tumours. Neuroblastoma are characterised by high cellularity, which translates into a high-signal and superior tumour-to-background contrast as a proxy of lesion conspicuity on DWI [10]. Just as tissue microstructure varies between different types of neuroblastic tumours with positive correlation of cellularity and malignant potential [14], so does the degree of restricted diffusivity as quantified by the ADC. Consequently, available studies consistently reported significant differences in ADC between NB and GN, with intermediate ADC values for GNB as an entity of mixed histological composition and of variable malignant potential. Gahr et al. [9] and Serin et al. [8] found some overlap between groups of malignant and benign disease in cohorts of 16 and of 24 patients, respectively. In contrast, another study on 29 patients with neuroblastic tumours, including four GBN [10] reported no overlap. A recently published report on 25 children with thoracoabdominal neuroblastic tumours examined with DWI at 3 T further refined analysis and correlated ADC with various histological subtypes and differentiation grades [15]. The overall picture is that of a continuous spectrum of disease, rather than of discrete categories, both on the histopathological level and on diagnostic imaging.
The results of our present study are in line with earlier findings in this respect. The ADC cutoff value of 1.05 identified on ROC analysis for differentiating malignant from benign disease falls exactly within the range proposed in an earlier study that used hardware of the same manufacturer, but analysed DWI scans acquired at 1.5 Tesla with somewhat different scan parameters and different techniques of ROI measurement [10].
A new aspect from our study is the use of ADC as an indicator of therapy response. The most commonly used quantitative parameter for measuring tumour response to therapy in a clinical setting is tumour size, or tumour volume. Yoo et al. [16] state that a therapy-induced decrease in tumour volume by > 40% is a strong predictor of event-free 5-year-survival in patients with high-risk neuroblastoma. In our study, tumour volume significantly decreased in all intermediate/high-risk NB/GNB patients with therapy. However, five of the seven patients in this group eventually suffered tumour recurrence in spite of tumour shrinkage of 48% and higher between baseline and follow-up scan. Interestingly, three of these patients with early relapse were seen with falling ADC values at follow-up, while ADC tended to increase at follow-up in low-risk NB patients under observation. Therefore, changes in tumour ADC measured during, or after, chemotherapy may outperform volume quantification as a non-invasive predictor of therapy response, similar to what has been reported from other malignancies in adult patients [17].
With regard to ADC as a possible predictor of event-free survival, baseline ADC < 0.80 × 10−3 mm2/s is as sensitive as a biomarker as the clinical risk group classification and carries a similar high negative predictive value, based on our data. We observed no tumour recurrence in our NB patients during the available observation period if baseline ADC were 0.80 × 10−3 mm2/s or higher. Such association between ADC and disease-free survival has, to our best knowledge, not been reported in paediatric tumours before. However, there is supportive data from adult cohorts, where in a head-to-head comparison of DWI and positron emission tomography/computed tomography (PET/CT), ADC performed as well as maximum standardised uptake value in predicting disease-free survival in patients with head and neck squamous cell carcinoma [18]. In a paediatric diagnostic setting, one would, of course, prefer MRI to PET/CT for risk stratification if both methods performed equally well.
Our study suffers methodological limitations mainly arising from the retrospective study design and from the small total number of patients enrolled. Although counting among the paediatric tumours with the highest incidence, neuroblastoma is nevertheless rare with a population-based age-adjusted incidence rate of 1.3 per 100.000 in Germany [19]. The number of patients treated in any single centre over a couple of years is too small to accumulate into a major cohort. Small sample size translates into a low power of statistical tests. For instance, Kaplan-Meier analysis of a test cohort and a validation cohort, as well as a Cox regression analysis, were not considered feasible in this small sample. Our findings are therefore to be considered very preliminary and are best evaluated in the context of other studies published on other small cohorts. Despite all differences in technical setup and analytic techniques, compared to earlier research, our results by and large support and underline existing knowledge on DWI of NB and GN. As in other studies, GNB constitutes the smallest, and the most volatile, subgroup. While the only GNB patient in our cohort exhibited some imaging features of benign disease, the clinical presentation with synchronous lymph node metastasis and subsequent tumour recurrence was strikingly malignant. The yet unclarified relation between biological potential of GNB and their features on diagnostic imaging and on histopathology invites further investigations.
Quantitative data collected by more than one observer or repeated analysis by the same reader would have allowed for evaluation of interobserver and intraobserver variability, as reported in an earlier publication [10], but were not available in this study. Our readings were performed by one experienced paediatric radiologist with the knowledge that previously published data demonstrated low variability in ADC quantification, even with readers at different levels of experience [10].
In conclusion, based on the new preliminary evidence from our study, we hold that DWI has already become a valuable and reliable imaging tool and may develop into a new risk biomarker for the diagnostic work-up of patients with neuroblastic tumours. Low ADC at first presentation is a highly characteristic of malignancy. Low, and falling, ADC with chemotherapy seems to predict early relapse, and perhaps poor outcome, in spite of therapy-induced reduction in tumour volume. Finally, in the light of ongoing discussions concerning the long-term safety of gadolinium-based contrast agents [20], DWI may help to avoid contrast agent administration, if scanned as a substitute, not as a supplement, to contrast-enhanced sequences, for instance at follow-up. Available data on image quality, lesion conspicuity and diagnostic performance seem to support such considerations [10], although verification of these preliminary findings through further studies would be welcome.