Synthetic MRI Combined With Clinicopathological Characteristics for Pretreatment Prediction of Chemoradiotherapy Response in Advanced Nasopharyngeal Carcinoma

Abstract

Objective

To explore the feasibility of synthetic magnetic resonance imaging (syMRI) combined with clinicopathological characteristics for the pre-treatment prediction of chemoradiotherapy (CRT) response in advanced nasopharyngeal carcinoma (ANPC).

Materials and Methods

Patients with ANPC treated with CRT between September 2020 and June 2022 were retrospectively enrolled and categorized into response group (RG, n = 95) and non RGs (NRG, n = 32) based on the Response Evaluation Criteria in Solid Tumors (RECIST) 1.1. The quantitative parameters from pre-treatment syMRI (longitudinal [T1] and transverse [T2] relaxation times and proton density [PD]), diffusion-weighted imaging (apparent diffusion coefficient [ADC]), and clinicopathological characteristics were compared between RG and NRG. Logistic regression analysis was applied to identify parameters independently associated with CRT response and to construct a multivariable model. The areas under the receiver-operating characteristic curve (AUC) for various diagnostic approaches were compared using the DeLong test.

Results

The T1, T2, and PD values in the NRG were significantly lower than those in the RG (all P < 0.05), whereas no significant difference was observed in the ADC values between these two groups. Clinicopathological characteristics (Epstein–Barr virus [EBV]-DNA level, lymph node extranodal extension, clinical stage, and Ki-67 expression) exhibited significant differences between the two groups. Logistic regression analysis showed that T1, PD, EBV-DNA level, clinical stage, and Ki-67 expression had significant independent relationships with CRT response (all P < 0.05). The multivariable model incorporating these five variables yielded AUC, sensitivity, and specificity values of 0.974, 93.8% (30/32), and 91.6% (87/95), respectively.

Conclusion

SyMRI may be used for the pretreatment prediction of CRT response in ANPC. The multivariable model incorporating syMRI quantitative parameters and clinicopathological characteristics, which were independently associated with CRT response, may be a new tool for the pretreatment prediction of CRT response.

INTRODUCTION

Nasopharyngeal carcinoma (NPC) is a prevalent malignant head and neck tumor with a high incidence in Southeast Asia and Southern China [1]. Owing to its deep position and asymptomatic nature in the early stages, more than 50% of patients present with advanced stage disease [2]. Although patients with advanced NPC (ANPC) can achieve significant survival benefits from chemoradiotherapy (CRT), some (10%–30%) experience local residual disease and relapse [3]. Therefore, it would be valuable to identify patients with CRT-insensitive ANPC before treatment, contributing to the avoidance of unnecessary CRT exposure and early adjustment of therapeutic strategies.

Magnetic resonance imaging (MRI) is a noninvasive technique that provides excellent soft-tissue contrast. Morphological information from conventional MRI has been used to evaluate therapeutic effects based on the Response Evaluation Criteria in Solid Tumors (RECIST) standard [4]. However, the application of RECIST is hysteretic because of its retrospective nature [5]. Although diffusion-weighted imaging (DWI) can reveal tissue microscopic properties and has been used in the evaluation of CRT response in patients with ANPC [6,7], DWI techniques based on echo-planar imaging are prone to image distortion that can affect diagnostic performance [8]. Therefore, a more effective MRI technique is necessary for the pretreatment prediction of CRT response in ANPC.

Longitudinal relaxation time (T1), transverse relaxation time (T2), and proton density (PD) are inherent and independent features of tissues in MRI that can reflect microstructural changes within lesions [9,10]. Synthetic MRI (syMRI) simultaneously provides T1, T2, and PD maps after one acquisition scan [11,12]. To date, syMRI has been used to differentiate benign and malignant tumors [13], evaluate tumor grade [14,15], identify indistinct tumor subtypes [16], and predict disease progression [17]. However, the clinical value of syMRI for the pretreatment prediction of CRT response in ANPC remains unclear.

In this study, we aimed to explore the potential use of quantitative parameters from syMRI (T1, T2, and PD) and DWI (apparent diffusion coefficient [ADC]) for the pretreatment prediction of CRT response in ANPC. The syMRI parameters and clinicopathological characteristics were further incorporated into a multivariable model to evaluate the performance of CRT response prediction in ANPC.

MATERIALS AND METHODS

Patient Selection

This study was approved by the Institutional Review Board of the Affiliated Hospital of Jiangnan University (IRB No. LS2020092). Informed consent was waived owing to the retrospective nature of the study. The flowchart of patient selection is shown in Figure 1. Between September 2020 and June 2022, 152 consecutive patients with biopsy-confirmed ANPC were enrolled. The inclusion criteria were as follows: 1) newly diagnosed NPC at clinical stage III or IV, 2) age >18 years, and 3) Karnofsky performance status score ≥80. The exclusion criteria were as follows: 1) no CRT or other/additional treatments (n = 11), 2) no syMRI (n = 3), 3) MRI artifacts (n = 6), and 4) no measurable lesion on MRI (n = 5). Finally, 127 patients with ANPC were selected for this study. Data on clinical factors (age, sex, serum lactate dehydrogenase [LDH] level, hemoglobin level, and Epstein–Barr virus [EBV]-DNA level) were collected for all patients with ANPC.

Fig. 1. Flowchart of patient selection. NPC = nasopharyngeal carcinoma.

MRI Acquisition

In the week before and 1 month after CRT, all patients underwent MRI examinations using a 3T MRI system (SIGNA™ Architect; GE HealthCare, Chicago, IL, USA) with a 28-channel phased-array coil. SyMRI was performed using the vender-provided sequence (MRI compilation) with the following parameters: repetition time (TR), 4000 ms; echo time (TE), 14.2/92.1 ms; slice thickness/gap, 4/0.4 mm; field of view (FOV), 24 × 24 cm²; matrix size, 320 × 256; number of excitations, 1; echo-train length, 16; and acquisition time, 3 minutes and 28 seconds. DW images were obtained using a single-shot echo-planar sequence, and the scan parameters were as follows: TR, 3457 ms; TE, 1.0 ms; slice thickness, 4 mm; FOV, 240 mm²; acquisition matrix, 128 × 130; and acquisition time, 2 minutes and 22 seconds. Two gradient factors (b = 0 and 1000 s/mm²) were used to determine the ADC map. Subsequently, contrast-enhanced T1-weighted imaging (CE-T1WI) was performed after gadolinium diethylenetriamine penta-acetic acid (0.469 g/mL) injection at a dose of 0.2 mL/kg. The scan parameters for CE-T1WI were as follows: TR, 431 ms; TE, 10 ms; slice thickness, 3 mm; slice gap, 1 mm; FOV, 24 × 24 cm²; number of excitations, 2; matrix size, 320 × 224; and acquisition time, 3 minutes and 37 seconds.

Imaging Analysis

All MR images were independently analyzed by two radiologists with 11 and 25 years of clinical experience who were blinded to the pathology reports and clinical data. All image analyses were performed for the primary tumor. The syMRI data were transferred into postprocessing software (SyMRI 7.2; SyntheticMR, Linköping, Sweden), and qualitative contrast-weighted images (including synthetic T1WI and T2WI) and quantitative relaxometry maps (T1, T2, and PD) were obtained. The region of interest (ROI) was manually drawn along the contour of the tumor on the slice with the largest tumor diameter on synthetic T2WI, with reference to CE-T1WI, to avoid obvious necrotic or cystic regions and peritumoral edema. Quantitative syMRI parameters (T1, T2, and PD) were obtained. The ROI on the ADC map was also drawn on the slice with the largest tumor diameter, and the mean ADC value was obtained. The quantitative parameters measured by the two radiologists were used to evaluate interobserver agreement and were averaged for further analysis. To evaluate intraobserver agreement, all syMRI parameters were remeasured by a radiologist with 11 years of clinical experience, with a minimum 1-month washout period.

The presence of extranodal extension (ENE), tumor volume, tumor stage, and tumor size were assessed by a radiologist with 11 years of clinical experience, who was blinded to the clinical and histopathological data. ENE was defined as a lymph node with infiltration of surrounding fat and/or muscle/skin/glands on synthetic T1WI, T2WI, or CE-T1WI. Tumor volume was calculated by multiplying the sum of all tumor areas on each slice by the image-reconstruction interval. Tumor stages were determined based on the 8th edition of the Union for International Cancer Control/American Joint Committee on Cancer tumor-node-metastasis system, including T stage, N stage, and clinical stages III (T0-2N2M0 or T3N0-2M0) and IV (T0-3N3M0 or T4N0-3M0 or T0-4N0-3M1) [18,19]. Tumor size was measured on a slice with the maximum diameter on CE-T1WI. Changes in tumor size after CRT were defined as follows: Change (%) = (Tumor size_{(before CRT)} - Tumor size_{(after CRT)})/Tumor size_{(before CRT)} × 100%. For patients with a complete response, the change in tumor size was 100%.

Treatment Procedure and Evaluation

All patients received two-cycle induction chemotherapy (21 days per cycle) consisting of 135 mg/m² paclitaxel on day 1 and 80 mg/m² nedaplatin on days 1–3. Two weeks after induction chemotherapy, all patients received concurrent CRT, including intensity-modulated radiation therapy at a total dose of 70–76 Gy/30–33 times/44–55 days, and two cycles of chemotherapy (21 days per cycle) using 75 mg/m² cisplatin on days 1–5 [20].

One month after CRT completion, MRI of the nasopharynx was analyzed by two radiologists with 11 and 5 years of clinical experience in the head and neck regions, respectively. CRT response was graded based on the RECIST version 1.1 standard. The criteria were as follows: complete response, disappearance of all target tumors; partial response, ≥30% decrease in the longest diameter; progressive disease, ≥20% increase in the maximum diameter or appearance of new lesions; and stable disease, between partial response and progressive disease [5]. Patients with complete response (n = 73) and partial response (n = 22) were categorized into the response group (RG, n = 95), and patients with stable disease (n = 27) and progressive disease (n = 5) were categorized into the non RG (NRG, n = 32) [21]. If the two radiologists performed different evaluations, a senior radiologist with more than 30 years of experience evaluated the data and made a final decision.

Histopathological Assessment

Before CRT, tissue samples were obtained via biopsy, which was performed after pretreatment MRI. Histopathological assessment of all patients was performed by a pathologist with 15 years of experience in head and neck tumors. Tumors were classified into keratinizing and non-keratinizing subtypes based on the presence of keratin on the tumor cell surface according to hematoxylin and eosin staining. Ki-67 expression was scored from 0 to 4 according to the staining intensity and cell ratio: 0, negative or weak staining in <10% of cells; 1, weak staining in 11%–30% of cells; 2, weak staining in >30% of cells or moderate staining in <30% of cells; 3, moderate staining in 30%–60% of cells; and 4, moderate or strong staining in >60% of cells. Scores of 0 and 1 represented low Ki-67 expression, whereas scores of 2, 3, and 4 represented high Ki-67 expression [22].

Statistical Analysis

The intraclass correlation coefficient (ICC) was used to assess inter- and intra-observer agreement, and ICC values <0.5, 0.5–0.75, 0.75–0.9, and >0.90 were indicative of poor, moderate, good, and excellent agreement, respectively [23]. The normality of the distribution was evaluated using the kurtosis-skewness test and quantile-quantile plot. Comparisons between the RG and NRG were analyzed using Student’s t-test for noncategorical data and Fisher’s exact test for categorical data. T1, T2, and PD values were compared pre- and post-treatment in patients with partial response, stable disease, and progressive disease, using paired-sample t-tests. Uni- and multivariable logistic regression analyses were performed to select independent factors for CRT response prediction. The selected independent factors were then used to construct a multivariable model using a logistic regression analysis. The Hosmer–Lemeshow test with bootstrapping was used to assess goodness of fit, with P > 0.05 indicating a good logistic regression model fit. The receiver-operating characteristic (ROC) curve was used to evaluate predictive performance. The areas under the ROC curve (AUC) for combinations of independent predictive factors were compared using the DeLong test. The optimal threshold, sensitivity, and specificity were determined by assessing the maximum Youden indices of ROC curves. All statistical analyses were performed using MedCalc (version 15.2.2; Mariakerke, Belgium) and SPSS (version 22.0; IBM Corp., Armonk, NY, USA). In this study, P < 0.05 was considered statistically significant. Given the exploratory nature of this study, we did not adjust for multiple comparisons.

RESULTS

This study involved 127 patients with ANPC, including 88 males and 39 females. As shown in Table 1, there were no significant differences between the RG and NRG in terms of age, sex, hemoglobin level, LDH level, tumor volume, initial tumor size, T stage, N stage, or pathological subtype. However, significant differences were found in EBV-DNA levels, ENE, clinical stage, and Ki-67 expression between the two groups.

Table 1. Clinicopathological characteristics in ANPC patients.

Characteristics		RG (n = 95)	NRG (n = 32)	P
Age, yr		53 ± 12	49 ± 14	0.120
Sex				0.659
	Male	67 (70.5)	21 (65.6)
	Female	28 (29.5)	11 (34.4)
Hemoglobin, g/L		138.89 ± 17.63	144.68 ± 13.77	0.093
LDH, U/L		196.00 ± 41.01	206.32 ± 40.39	0.219
EBV-DNA				0.012
	<4000	61 (64.2)	12 (37.5)
	≥4000	34 (35.8)	20 (62.5)
Extranodal extension				0.004
	Yes	31 (32.7)	20 (62.5)
	No	64 (67.3)	12 (37.5)
Tumor volume, cm3		88.49 ± 97.16	99.22 ± 46.38	0.549
Tumor size, cm				0.419
	<3	50 (52.6)	14 (43.8)
	≥3	45 (47.4)	18 (56.3)
T stage				0.597
	T1	2 (2.1)	1 (3.1)
	T2	34 (35.8)	11 (34.4)
	T3	47 (49.5)	13 (40.6)
	T4	12 (12.6)	7 (21.9)
N stage				0.604
	N0	9 (9.5)	2 (6.3)
	N1	15 (15.8)	7 (21.9)
	N2	54 (56.8)	15 (46.8)
	N3	17 (17.9)	8 (25.0)
Clinical stage				0.006
	III	64 (67.4)	11 (34.4)
	IV	31 (32.6)	21 (65.6)
Pathological type				0.340
	Keratinizing	9 (9.5)	4 (12.5)
	Nonkeratinizing	86 (90.5)	28 (87.5)
Ki-67 expression				0.002
	Low	68 (71.6)	13 (40.6)
	High	27 (28.4)	19 (59.4)

Data are mean ± standard deviation or number of patients (%).

ANCP = advanced nasopharyngeal carcinoma, RG = response group, NRG = non-response group, LDH = lactate dehydrogenase, EBV = Epstein–Barr virus

Intra- and interobserver agreements were good to excellent for the syMRI parameters and ADC, with ICCs ranging from 0.884–0.901 and 0.866–0.882, respectively (Supplementary Table 1). Compared with RG, NRG had significantly lower T1, T2, and PD values (Fig. 2, Table 2), whereas there was no significant difference in ADC values between the NRG and RG (Fig. 2, Table 2). In comparison with posttreatment lesions, pretreatment lesions had significantly lower T1, T2, and PD values (Supplementary Fig. 1, Supplementary Table 2). Representative MR images of the ANPC in the RG and NRG are shown in Figures. 3, 4.

Fig. 2. Comparison of syMRI quantitative parameters and ADC values between the RG and NRG. A-D: Compared with the RG, the NRG had significantly lower T1 (A), T2 (B), and PD (C) values (A-C), whereas there was no significant difference in ADC values between these two groups (D). syMRI = synthetic magnetic resonance imaging, ADC = apparent diffusion coefficient, RG = response group, NRG = nonresponse group, T1 = longitudinal relaxation time, T2 = transverse relaxation time, PD = proton density.

Table 2. Comparison results of syMRI quantitative parameters and ADC value between RG and NRG.

	T1, ms	T2, ms	PD, pu	ADC, ×10-6 mm2/s
RG	1281.14 ± 114.84	82.95 ± 9.02	86.35 ± 4.69	774.89 ± 139.27
NRG	1161.62 ± 138.72	78.10 ± 10.44	81.00 ± 4.25	832.09 ± 100.59
P	<0.001	0.046	<0.001	0.085

Data are mean ± standard deviation.

syMRI = synthetic magnetic resonance imaging, ADC = apparent diffusion coefficient, RG = response group, NRG = non-response group, T1 = longitudinal relaxation time, T2 = transverse relaxation time, PD = proton density, pu = percentage unit

Fig. 3. A representative 50-year-old male patient with ANPC (stage IV, T4N2M0) and stable disease in the NRG. A-D: The ROI was manually delineated, as shown by the red/black contours. In comparison with RG, relaxation maps showed significantly lower T1, T2, and PD values (T1, 1116 ms; T2, 72 ms; PD, 79.2 pu), whereas there was no significant difference in ADC values between RG and NRG. E: CE-T1WI showed that the tumor was 23.04 mm in diameter before CRT. F: After CRT, CE-T1WI demonstrated that the tumor was 22.74 mm in diameter. ANPC = advanced nasopharyngeal carcinoma, T = tumor, N = node, M = metastasis, NRG = nonresponse group, ROI = region of interest, RG = response group, T1 = longitudinal relaxation time, T2 = transverse relaxation time, PD = proton density, pu = percentage unit, ADC = apparent diffusion coefficient, CE-T1WI = contrast-enhanced T1-weighted imaging, CRT = chemoradiotherapy.

Fig. 4. A representative 46-year-old female patient with ANPC (stage III, T3N2M0) with complete response in the RG. A-D: The ROI was manually delineated, as shown by the red/black contours. In comparison with the NRG, the relaxation maps showed significantly higher T1, T2, and PD values (T1, 1241 ms; T2, 92 ms; PD, 86.5 pu), whereas there was no significant difference in the ADC values between the RG and NRG. E: CE-T1WI demonstrated that the tumor was 26.74 mm in diameter before CRT. F: After CRT, CE-T1WI shows that the tumor has almost disappeared. ANPC = advanced nasopharyngeal carcinoma, T = tumor, N = node, M = metastasis, RG = response group, ROI = region of interest, NRG = nonresponse group, T1 = longitudinal relaxation time, T2 = transverse relaxation time, PD = proton density, pu = percentage unit, ADC = apparent diffusion coefficient, CE-T1WI = contrast-enhanced T1-weighted imaging, CRT = chemoradiotherapy.

As summarized in Table 3, T1, PD, EBV-DNA level, clinical stage, and Ki-67 expression had significant independent relationships with CRT response in the multivariable logistic regression analysis.

Table 3. Uni- and multivariable logistic regression analysis in the prediction of non-response group in ANPC patients.

Predictors		Univariable analysis			Multivariable analysis*
		OR	95% CI	P	OR	95% CI	P
T1, ms		0.99	0.98–1.00	<0.001	0.99	0.98–1.00	0.031
T2, ms		0.94	0.89–0.99	0.046	1.11	0.98–1.25	0.097
PD, pu		0.67	0.54–0.84	<0.001	0.65	0.47–0.91	0.011
ADC, ×10-6 mm2/s		1.00	0.99–1.00	0.092
EBV-DNA
	<4000	Reference			Reference
	≥4000	4.28	1.51–12.18	0.006	25.21	2.73–231.90	0.004
Extranodal extension
	No	Reference			Reference
	Yes	4.68	1.66–13.24	0.004	3.95	0.68–22.86	0.125
Clinical stage
	III	Reference			Reference
	IV	5.53	1.93–15.82	0.002	6.72	1.11–40.63	0.037
Ki-67 expression
	Low	Reference			Reference
	High	10.88	3.42–36.62	<0.001	7.39	1.04–50.32	0.045

^*Variable with P < 0.05 at univariable analysis were included in the multivariable analysis.

ANPC = advanced nasopharyngeal carcinoma, OR = odds ratio, CI = confidence interval, T1 = longitudinal relaxation time, T2 = transverse relaxation time, PD = proton density, pu = percentage unit, ADC = apparent diffusion coefficient, EBV = Epstein–Barr virus

In addition, as shown in Table 4 and Supplementary Figure 2, the AUCs for T1, T2, and PD ranged from 0.615 to 0.818, and the AUCs for T1 and PD were significantly higher than those for T2 (T1 vs. T2, P = 0.004; PD vs. T2, P = 0.003), whereas no significant difference was found between the AUCs for T1 and PD (T1 vs. PD, P = 0.950). We also investigated the possibility of improving predictive performance by combining T1 + PD, T1 + T2, T2 + PD, and T1 + T2 + PD. These results demonstrated some improvement in prediction efficacy. A multivariable model incorporating all independent predictive factors (T1, PD, EBV-DNA level, clinical stage, and Ki-67 expression) showed excellent predictive performance for CRT response in patients with ANPC (Table 4, Supplementary Fig. 2). The formula of the multivariable model was as follows: Logit (P) = -0.006 × T1 - 0.429 × PD + 3.015 × EBV-DNA + 2.113 × clinical stage + 1.920 × Ki-67 + 30.643. The Hosmer–Lemeshow test showed that this model had good logistic regression model fit (P = 0.775). The AUC for the multivariable model was significantly higher than that for T1 + T2 + PD (P = 0.029).

Table 4. Diagnostic performance in pretreatment prediction of CRT response in ANPC patients.

	AUC (95% CI)	Threshold	Youden index	Sensitivity, %	Specificity, %
T1, ms	0.815 (0.717–0.890)	1190	0.643	81.3 (26/32)	83.2 (79/95)
T2, ms	0.615 (0.505–0.717)	75.0	0.325	46.9 (15/32)	85.3 (81/95)
PD, pu	0.818 (0.721–0.893)	83.1	0.504	75.0 (24/32)	74.7 (71/95)
T1 + PD*	0.869 (0.780–0.932)	0.28	0.706	84.4 (27/32)	85.3 (81/95)
T1 + T2*	0.831 (0.736–0.903)	0.28	0.658	84.4 (27/32)	83.2 (79/95)
T2 + PD*	0.821 (0.724–0.895)	0.27	0.520	75.0 (24/32)	75.8 (72/95)
T1 + T2 + PD*	0.897 (0.813–0.952)	0.39	0.721	84.4 (27/32)	86.3 (82/95)
Multivariable model*	0.974 (0.915–0.996)	0.23	0.876	93.8 (30/32)	91.6 (87/95)

^*The formula for prediction was as follow: T1 + PD: Logit (P) = -0.005 × T1 - 0.294 × PD + 29.723. T1 + T2: Logit (P) = -0.009 × T1 - 0.005 × T2 + 10.946. T2 + PD: Logit (P) = -0.006 × T2 - 0.404 × PD + 32.032. T1 + T2 + PD: Logit (P) = -0.006 × T1 -0.039 × T2 - 0.316 × PD + 29.903. Multivariable model: Logit (P) = -0.006 × T1 - 0.429 × PD + 3.015 × EBV-DNA + 2.113 × clinical stage + 1.920 × Ki-67 + 30.643.

CRT = chemoradiotherapy, ANPC = advanced nasopharyngeal carcinoma, T1 = longitudinal relaxation time, PD = proton density, T2 = transverse relaxation time, EBV = Epstein–Barr virus, AUC = area under the receiver-operating characteristic curve, CI = confidence interval, pu = percentage unit

DISCUSSION

In this study, we investigated the use of syMRI quantitative parameters and their combination with clinicopathological characteristics in the pretreatment prediction of CRT response in patients with ANPC. The results demonstrated that syMRI quantitative parameters (T1, T2, and PD) in the NRG were significantly lower than those in the RG. The multivariable model incorporating independent predictive factors (T1, PD, EBV-DNA level, clinical stage, and Ki-67 expression) exhibited excellent predictive performance.

Prior to CRT, the clinicopathological characteristics of patients with ANPC were evaluated. Our results demonstrated that the clinical stage and Ki-67 expression were significantly different between RG and NRG, which is consistent with previous studies on ANPC [21,22]. These similar findings may be attributed to the fact that advanced stage tumors and those with high cell proliferation may often be accompanied by necrosis due to ischemia and hypoxia, thereby giving rise to CRT resistance [21,24]. Furthermore, consistent with previous studies [1,25], higher EBV-DNA levels were found in the NRG than in the RG. The reason behind this may be that hypoxic tumors have more necrosis, thereby shedding more EBV-DNA fragments into the bloodstream [26]. Additionally, in line with a previous study on head and neck tumors [27], ENE was more frequent in the NRG than in the RG. This may be because tumor cells cannot be eradicated in patients with ENE because they can easily gain access to blood circulation [28]. Therefore, clinical stage, Ki-67 expression, EBV-DNA level, and ENE may be useful factors for predicting CRT response in ANPC.

As previously reported, hypoxia is a common phenomenon in solid tumors that contributes to CRT resistance owing to poor radiosensitivity and limited chemotherapeutic drug delivery to the inner core of solid tumors [29]. Baidya Kayal et al. [30] found that the T1 value in the NRG was significantly lower than that in the RG for osteosarcoma. In the present study, a similar result was found regarding the T1 value in the two groups of patients with ANPC. The reason for this may be that tumors with CRT resistance contain high proportions of hypoxic cells, which can decrease the T1 value owing to concomitant necrosis and the release of macromolecules, paramagnetic ions, and proteins [31]. Hence, a lower T1 value may be indicative of CRT resistance in patients with ANPC. In addition, our study found a significantly lower T2 value in the NRG than in the RG, which is consistent with the results of a previous study on rectal tumors [32]. This may be because hypoxic tumors can decrease their T2 values through the release of macromolecules, paramagnetic ions, and proteins. However, our study showed that the T2 value did not have a significant independent relationship with CRT response. A possible explanation for this finding could be that, although hypoxic cells can shorten T2, edema may increase the ratio of free to bound water, thereby increasing T2 during the development of necrosis. These potentially competing processes may preclude a significant independent relationship between T2 and CRT response [31]; however, this hypothesis should be confirmed in future studies. Consistent with previously reported findings [17], our study showed that the PD value was significantly lower in the NRG than in the RG. The underlying mechanism may be that responsive tumors have high blood volumes and microcirculatory perfusion, which can increase the amount of extracellular water and elevate the PD value [21,33]. Consequently, the PD value may be another useful parameter for the pretreatment prediction of CRT response in ANPC. Furthermore, this study found that the multivariable model incorporating syMRI parameters and clinicopathological factors further improved predictive performance. Thus, syMRI may be helpful for the pretreatment prediction of CRT response in ANPC.

In this study, there was no significant difference in ADC values between the NRG and RG, which is in line with previous studies on NPC [34,35]. However, other studies have suggested that the ADC value is a predictive factor for tumor response in patients with NPC [36]. This discrepancy might be explained by variations in patients’ clinical factors, MRI characteristics and techniques, and timing of therapeutic response evaluation.

In this study, a multidynamic multiecho-based syMRI technique was used. There are other syMRI techniques that quantify the relaxometric parameters (T1, T2, and PD), such as multipathway, multiecho (MPME) imaging, and MR fingerprinting (MRF) [37,38]. However, MPME-based quantitative maps have relatively high noise, which limits their voxel-based interpretation [39]. Although MRF quantitative maps may achieve excellent isotropic resolution, the repeatability, reproducibility, and validation of this technique must be confirmed before it can be a reliable clinical tool [40]. Additionally, there are other quantitative methods for generating T1 and T2 maps; however, these methods are not capable of synthesizing contrast-weighted images [41]. In addition, previous studies have reported that patients who achieved early tumor regression had better prognoses, whereas the presence of early residual tumors was an independent negative prognostic factor [42,43]. Therefore, the response assessment in the present study was timed to occur during the early stages of CRT completion.

This study has some limitations. First, this retrospective study was conducted at a single center and involved a small sample of patients with ANPC. Therefore, a multicenter study with a large cohort of patients with ANPC is required to validate the generalizability and clinical implications of this study. Second, RECIST was used as the gold standard for evaluating the treatment response to CRT, which should be verified by tissue diagnosis in future studies. Third, the single-slice measurement of syMRI parameters may be influenced by the reader and may only represent a portion of the tumor. Whole-tumor measurement more accurately reflects overall tumor characteristics and allows for the assessment of tumor heterogeneity, which should be performed in future studies. Fourth, this study lacks external validation. Fifth, ENE was assessed via imaging in this study, which is less accurate than pathological assessment for predicting outcomes in head and neck cancer.

In conclusion, syMRI quantitative parameters (T1, T2, and PD) were useful for the pretreatment prediction of CRT response in patients with ANPC. The multivariable model incorporating syMRI parameters and clinicopathological factors, which were independently associated with CRT response, may serve as a new tool for the pretreatment prediction of CRT response.

Availability of Data and Material

The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.

Supplement

The Supplement is available with this article at https://doi.org/10.3348/kjr.2024.0385.