Among types of head and neck squamous cell carcinoma (HNSCC), nasopharyngeal carcinoma (NPC) is the most common and is highly prevalent in Southeast Asia and Southern China [1]. The majority of NPC was found to be non-keratinizing in endemic regions [2] and was highly sensitive to irradiation and chemotherapeutic drugs. The primary treatment regimen for NPCs is radiotherapy (RT) coupled with or without chemotherapy. However, there are limited effective non-invasive image modalities for the early prediction of tumor response to chemoradiotherapy (CRT).

Intensity-modulated radiation therapy (IMRT) for NPC has been reported to provide satisfactory outcomes compared to other RT techniques [3,4]. For advanced diseases, CRT shows a better prognosis than RT alone, especially sequential CRT and concurrent chemoradiotherapy (CCRT) [5-7]. In particular, studies reported higher survival rates in patients undergoing CCRT than sequential CRT [6,7]. Notably, radio-resistance, tumor recurrence, and the development of distant metastases remain the contributing reasons for treatment failure. In this regard, an optimal imaging regimen is critical for early assessment of radiosensitivity of tumor. A valid imaging modality was explored to assess the treatment responses of NPCs.

Routine magnetic resonance imaging (MRI) currently provides a better soft tissue contrast and spatial resolution in evaluating the extent of primary nasopharynx tumor and positive cervical nodes and also plays a conventional role in assessing tumor’s response to CRT. However, morphological images only reveal macroscopic changes in tumor size that may not be apparent, and the initial treatment and characterization of immediate therapeutic effects could be limited. Recent studies primarily focused on the functional treatment response of tumors by various MRI techniques. For instance, diffusion-weight imaging (DWI) measures and characterizes the thermal motion of water molecules. In particular, the apparent diffusion coefficient (ADC) has been validated as a potential imaging biomarker to identify tumor treatment response [8]. However, ADC is calculated based on a monoexponential model that reflects the combined effects of diffusion and perfusion [9].

In contrast, intravoxel incoherent motion (IVIM), which was initially described by Le Bihan et al. in 1986 [9], is a bi-exponential model to separately quantify tissue diffusion and perfusion [10]. In recent decades, IVIM-DWI has gained attentions due to its effective diagnostic capability in tumors, including those of the head and neck [11,12]. Moreover, IVIM parametrics have been correlated with clinical staging [13] as well as the treatment response of HNSCC [14,15]. Our previous study validated the potential value of IVIM-DWI in predicting the early treatment response of neoadjuvant chemotherapy (NAC) in local to regionally advanced NPCs [16]. Nevertheless, the assessment of IVIM-DWI regarding the radio-sensitivity of NPCs is yet to be explored. Since early prediction of treatment response through IVIM-DWI might arise significant impacts on patient-care planning and treatment regimens, the objective of this study is to evaluate the clinical value of IVIM-DWI in predicting the treatment response to IMRT in NPCs.

The current study investigated the assessment of treatment responses in NPC by IVIM-DWI. IVIM parametrics including D, f, and D* altered significantly for responders during the course of IMRT. In particular, D and f were more sensitive to assess response in the early course of the treatment, whereas D* revealed changes in later stages.

The bi-exponential model of IVIM demonstrated the heterogeneity of water diffusion and differentiated the contributions from perfusion from diffusion [9,10,20]. Compared to conventional ADC measurement by DWI, IVIM parametrics can simultaneously and separately quantify tissue perfusion and diffusion, and better characterize the pathological features of tumor tissues as well as their response to chemoradiotherapy. Upon radiation therapy, large numbers of tumor cells underwent apoptosis, resulting in a higher degree of fluid migration into the extracellular space (i.e., pure diffusion). Meanwhile, the obstruction of microcirculation became severe, leading to less blood flow in capillary networks (i.e., pseudo-diffusion). Both the cellularity and immature microvessels in tumor tissues decreased dramatically. The overall effect of these pathological events led to a significant increase in D as well as decrease in D*, which were found in our data. Lai et al. [12] reported a similar diffusion perfusion change in residual NPC tumors after RT compared with post-radiation fibrosis. Furthermore, both D and D* were found to be significantly altered in responders relative to non-responders, indicating their potential in predicting the short-term treatment response of NPC [21-25].

Recent studies have employed diffusion MRI to monitor the treatment response of tumors in CRT [21-23]. Our previous study also demonstrated that NPC with a lower baseline D₀, which had a higher cellularity, responded better to NAC [16]. Because D composes diffusion in the extracellular and intracellular space, an alteration in D reflected the microstructure changes for the assessment of the treatment response in IMRT. This was consistent with Mardor et al.’s study [24] showing that both ADC and R_D (a diffusion index indicates the normalized summation of signal decay) before treatment were closely correlated with tumor radio-sensitivity. Although there were no significant changes in D after the fifth fraction, the D₅ of responders was significantly larger than that of non-responders. ROC analysis suggested that D₅ could differentiate treatment responses with AUC, sensitivity, and specificity of 0.773, 72.2%, and 90.9%, respectively. Lu et al. [25] also reported that HNSCC patients with a lower standard deviation of pre-treatment D responded better to treatment. Similar, studies also demonstrated the prognostic value of D [16,21-25]. However, perfusion measured by MRI has been used to assess the therapeutic effects and long-term prognosis for patients with malignancies. The perfusion level of tissues was suggested to potentially predict the tumor sensitivity of CRT [26,27]. Lai et al. [12,13] reported that f could effectively distinguish residual tumors of NPC from radiation-related fibrous tissues, as well as tumor stages. HNSCC patients with a smaller perfusion fraction were always accompanied with a low survival rate [22]. Interestingly, our present results showed that responders expressed a smaller baseline f₀, a larger f₁₀ and Δf₁₀ relative to non-responders. These may infer the irradiation damages on the integrity of microvascular walls in the early course of IMRT.

D and f were more sensitive in the early course of treatment, whereas D* changes more in later course in responders relative to non-responders. The ROC curve analysis further illustrated that D₅, D*₂₀, and f₁₀ better differentiate responders from non-responders. These findings may indicate that the destruction of cell structures by irradiations in radio-sensitive tumors was more predominant early in the treatment, whereas damage to microcirculation was more evident at the later course of radiation. Larocque et al. [28] and Koh [29] both demonstrated that early changes in diffusion parametrics could be used to evaluate the sensitivity of RT and/or chemotherapy. They observed that the average tumor ADC value increased significantly 7 days after irradiation [28] and that the correlation between the decrease in perfusion parametrics (D* and f) and tumor shrinkage 7 days after treatment significantly increased [29]. Hong et al. [8] also suggested ADC to predict the short-term treatment response to RT in NPC. All of the above supported that the IVIM parametrics reflect tumor intrinsic radiosensitivity and biological response.

Additionally, IVIM parametrics could indicate a response of various treatment regimens (i.e., IMRT with or without neoadjuvant or concurrent chemotherapy). D₀ of the NAC+IMRT group was significantly larger than IMRT alone and CCRT, whereas D*₀ of the NAC+IMRT group was significantly smaller than those of the latter. The D₂₀, f₂₀, and D*₂₅ parametrics between IMRT alone and NAC+IMRT groups were significantly different, whereas D*₂₅ was significantly different between IMRT alone and the CCRT groups. These indicated a significant increase in D and decrease in D* caused by NAC. This could be attributed to a decrease the cellularity and increase extracellular space upon NAC. Additionally, a decrease in tumor perfusion may suggest microcirculation blockage, which in turn influences the oxygen metabolism of tumor cells and the radio-sensitivity of NPC. No significant difference was found in IVIM parametrics between NAC+IMRT and CCRT throughout the course of treatment. This may suggest similar cytotoxic effects on tumor cells between neoadjuvant and concurrent chemotherapy.

To further solidify the evidence presented in this study, additional examinations will be assessed. First, histopathological examinations during radiations were not obtained for ethical reasons. Complementary studies on NPC xenografts are needed to analyze the pathological changes during fractional radiations. Second, the current study only evaluated local control and short-term treatment outcomes of primary NPC tumors, while initial staging (such as WHO grades and nodal or distant metastasis) was not considered between responders and non-responders. Nevertheless, the latest follow-up data indicated that three cases of local and regional relapse and two cases of distant metastases were all found in non-responders rather than responders. In this regard, a prospective study with a prolonged follow-up period to investigate the long-term survival of patients and cancer staging would be possible. Lastly, the sample size of this prospective study was relatively small, which should be expanded.

Compared to non-responders, responders presented significantly larger changes in D and f at the early course of treatment and D* later in the treatment. IVIM-DWI could effectively assess the short-term treatment response to RT and could be recommended to predict the radiosensitivity of NPCs.