Magnetic Resonance Imaging Features as Predictors of Response and Prognosis Following Neoadjuvant Therapy in Rectal Cancer: A Retrospective Cohort Study

Abstract

Background

MRI-detected circumferential resection margin (CRM) involvement, extramural venous invasion (EMVI), and tumor deposits (TDs) are established high-risk features associated with poor prognosis in rectal cancer. These features may help identify patients who are more likely to benefit from neoadjuvant therapy (NAT). However, the prognostic significance of changes in these MRI high-risk features after NAT remains unclear. The objective of this study was to determine whether baseline MRI high-risk features can identify patients likely to benefit from NAT and whether their changes after NAT predict treatment response and prognosis in rectal cancer.

Methods

This retrospective single-center cohort study included patients who underwent curative resection for rectal adenocarcinoma and received neoadjuvant therapy (NAT) at Sun Yat-sen University, Sixth Affiliated Hospital, between 2017 and 2019. Patients with MRI-detected CRM involvement, EMVI, or TDs were classified as MRI high-risk. Those who showed a shift from MRI high-risk to MRI low-risk—defined by the absence of CRM involvement, EMVI, or TDs—on post-treatment scans were considered good responders. The survival outcomes of good responders were then compared to those of poor responders, who remained persistently MRI high-risk. Additional analyses were performed within the baseline MRI high-risk subgroup to determine which MRI high-risk features were most strongly associated with prognosis.

Results

The study included 302 patients in total. Of these, 146 (48.3%) were classified as MRI high-risk and 156 (51.7%) as MRI low-risk based on pre-treatment imaging. The high-risk group had significantly worse outcomes: three-year disease-free survival (DFS) was 54.8% compared with 93.6% in the low-risk group (p < 0.001), three-year overall survival (OS) was 76% versus 98.1% (p < 0.001), and the rate of local recurrence (LR) at three years was 10.9% compared to 1.3% (p = 0.002). Among the 146 patients initially identified as MRI high-risk, those who converted to low-risk status after treatment showed improved outcomes, with a three-year DFS of 82.9%, OS of 94.3%, and LR rate of 2.9%. Within the baseline MRI high-risk subgroup, baseline mrEMVI and mrTD, as well as post-treatment ymrMRF, ymrEMVI, and ymrTD, were associated with worse DFS.

Conclusions

Baseline MRI high-risk features and their persistence after NAT were associated with poor prognosis in rectal cancer. MRI risk conversion after NAT may help identify a subgroup of initially high-risk patients with more favorable outcomes.

Keywords

rectal cancer magnetic resonance imaging (MRI)neoadjuvant therapy (NAT)circumferential resection margin (CRM)extramural venous invasion (EMVI)tumor deposits (TDs)prognosis

Introduction

Neoadjuvant therapy (NAT) is now commonly used in treating locally advanced rectal cancer to increase the chances of complete tumor removal and improve patient outcomes.^1,2 Despite its widespread use, the justification for applying NAT in all rectal cancer cases remains uncertain.^3-5 For example, the Dutch trial showed no significant difference in 10-year overall survival (OS) for patients with stage II disease who received NAT compared to those who did not (50% vs. 55%; p = 0.242).⁶ Likewise, a study by Cho et al found that patients with T3ab rectal cancer had similar five-year disease-free survival (DFS) whether they received preoperative chemoradiotherapy followed by surgery or had surgery alone (88% vs. 87%).⁷ In addition, NAT can lead to treatment-related complications and may increase the complexity of performing total mesorectal excision (TME).^8-10 These concerns highlight the need for better patient selection to determine who is most likely to benefit from NAT.

Multiple studies have shown that rectal cancer patients with MRI-detected circumferential resection margin (CRM) involvement, extramural venous invasion (EMVI), or tumor deposits (TDs) face a higher risk of both local and distant recurrence, along with reduced survival rates.^11-13 In research by Lord AC et al, patients were divided into high-risk and low-risk categories based on MRI findings, including CRM status, presence of EMVI, and TDs. Their results indicated that individuals in the high-risk group had significantly poorer five-year disease-free survival (DFS) and overall survival (OS), at 66% and 71% respectively, compared to 88% and 89% in the low-risk group.¹⁴ These MRI high-risk features may therefore help identify patients with biologically aggressive disease who are more appropriate candidates for NAT. In this context, MRI is not only a staging tool, but also a potentially valuable instrument for treatment selection.

Although the prognostic significance of these baseline MRI findings has been well established, their value in predicting treatment response and post-treatment prognosis following NAT has not been fully clarified. In practice, some patients with baseline MRI high-risk features show disappearance of these adverse findings after NAT, whereas others continue to have persistent MRI high-risk disease. Whether such dynamic changes in MRI-defined risk reflect treatment response and translate into differences in long-term oncologic outcomes remains unclear. Therefore, this study aimed to investigate whether baseline MRI high-risk features can help identify patients who may benefit from NAT and, more importantly, whether changes in these MRI risk features after NAT are associated with treatment response and prognosis in rectal cancer.

Methods

Study Design

We conducted a retrospective single-center cohort study of patients with rectal adenocarcinoma who received NAT followed by curative surgery at the Sixth Affiliated Hospital of Sun Yat-sen University between 2017 and 2019. Collected variables included demographic data and MRI findings. To protect patient privacy, all data were fully anonymized before use, and all patient details have been de-identified. This study was approved by the Institutional Review Board of The Sixth Affiliated Hospital, Sun Yat-sen University (approval number: 2022ZSLYEC-615; approval date: December 16, 2022), and conducted in accordance with the ethical standards of the institutional research committee and the Helsinki Declaration of 1975, as revised in 2024. The requirement for informed consent was waived by the approving IRB due to the retrospective study design and full anonymization of all patient clinical data. This study is reported in accordance with the STROBE statement.¹⁵ According to preliminary data, the expected 3-year DFS was approximately 89% in the MRI low-risk group and 68% in the MRI high-risk group. Using a two-sided alpha of 0.05, 80% power, and a 1:1 allocation ratio, the minimum required sample size was 59 patients per group (118 patients in total). After accounting for 10% incomplete data or loss to follow-up, the minimum required sample size was 132 patients.

Inclusion and Exclusion Criteria

All patients were consecutively identified from the institutional database during the predefined study period and then screened according to the prespecified inclusion and exclusion criteria. In order to be included in the study, patients needed to meet all the following inclusion criteria: 1) Underwent primary rectal adenocarcinoma resection; 2) Received baseline MRI before NAT and restaging MRI after NAT before surgery; 3) Treatment-naive patients with histological or cytological documentation of rectal adenocarcinoma (<12 cm from the anal verge); 4) Clinical stage of T3Nx or T1-3N+ at initial diagnosis; 5) Non-complicated primary tumor (without complete obstruction, perforation, bleeding); 6) Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1. The exclusion criteria included: 1) Treated with transanal local excision; 2) With distant metastases at the time of initial diagnosis; 3) Subjects with a history of a prior malignancy within the past 5 years, except for adequately treated basal cell or squamous cell skin cancer; 4) Subjects with a history of any arterial thrombotic event within the past 6 months. This includes angina (stable or unstable), myocardial infarction, transient ischemic attacks, or cerebrovascular accident. 5) Incomplete essential imaging, eligibility, or follow-up information.

MRI Features

To assess CRM status, the shortest distance between the tumor and the mesorectal fascia was measured on baseline MRI scans. On high-resolution MRI, CRM involvement was defined as the presence of tumor signal intensity at or within 1 mm of the mesorectal fascia.¹⁶ For low-lying tumors, involvement was defined as extension into the intersphincteric space or within 1 mm of the levator ani or external anal sphincter.¹⁷ EMVI was diagnosed when a continuous expansion of perirectal veins exhibited intermediate signal intensity consistent with tumor infiltration.^18,19 TDs were identified on MRI as irregular nodules within the mesorectum that disrupted venous structures but were discontinuous from the primary lesion.¹² TDs were differentiated from metastatic lymph nodes by their inseparability from adjacent veins on two orthogonal planes and their tapering configuration into the vessel lumen, producing a typical “comet-tail” morphology. Patients were classified as MRI high-risk if they had any of the following: CRM involvement, EMVI, or TDs. Those without these features were categorized as MRI low-risk. All patients underwent a restaging MRI (yMRI) within six weeks of completing NAT. Patients initially categorized as high-risk but reclassified as low-risk on post-treatment imaging were defined as good responders (Figure 1), while those who remained high-risk were considered poor responders.

Figure 1.

MRI features before and after NAT CRM involved (A), EMVI positive (B) and TD positive (C) on MRI at baseline. CRM clear (D), EMVI negative (E) and TD negative (F) on post-treatment scans

Radiographic TNM Staging

Radiographic TNM staging was defined as follows: T0 indicated no visible primary tumor; Tis referred to carcinoma in situ confined to the lamina propria without penetrating the muscularis mucosae; T1 represented invasion into the submucosa without reaching the muscularis propria; T2 denoted tumor invasion into the muscularis propria; T3 described extension beyond the muscularis propria into the perirectal tissues; and T4 indicated invasion of the visceral peritoneum or direct involvement of adjacent organs or structures. Regional lymph node staging was categorized as: N0 for no involved nodes; N1 for metastasis in one to three nodes; and N2 for metastasis in four or more nodes. Lymph nodes were considered positive based on size and morphology: nodes with a short axis <5 mm were labeled positive if they showed all three malignant features (indistinct borders, heterogeneous signal, and round shape); nodes measuring 5–9 mm were positive if they exhibited at least two of these features; and any node larger than 9 mm was considered positive regardless of appearance.^20,21 TNM downstaging was defined as a radiologic shift from T3–T4 to T0–T2, or from N1–2 to N0.^22,23

Treatment and Outcomes

Treatment strategies were tailored to each patient through thorough deliberation by a multidisciplinary team, adhering to protocols set forth by the National Health Commission of the People’s Republic of China. All patients underwent NAT before surgery. The systemic treatment backbone consisted of oxaliplatin-based doublet chemotherapy, administered as either FOLFOX (oxaliplatin, fluorouracil, and leucovorin) or XELOX (oxaliplatin plus capecitabine), according to institutional practice and patient tolerance. Neoadjuvant treatment was delivered using one of three predefined approaches: (1) six cycles of preoperative FOLFOX/XELOX alone; (2) short-course pelvic radiotherapy followed by chemotherapy; or (3) chemotherapy combined with concurrent long-course radiotherapy. In the short-course radiotherapy group, pelvic radiotherapy was delivered using volumetric modulated arc therapy (VMAT) at a total dose of 25 Gy in 5 fractions (5 Gy per fraction) before the initiation of chemotherapy, followed by six cycles of FOLFOX/XELOX. In the long-course radiotherapy group, pelvic radiotherapy was likewise delivered using VMAT at a total dose of 50 Gy in 25 fractions (2 Gy per fraction) during cycles 2–4 of chemotherapy. All patients underwent radical TME upon completion of the predefined NAT regimen. Postoperative adjuvant chemotherapy was generally recommended for patients with pathologic T3+ and/or N+ disease, most commonly with FOLFOX or XELOX, and was individualized according to final pathology and multidisciplinary team decision-making. This study evaluated follow-up endpoints including DFS, OS, and LR. DFS referred to the proportion of individuals remaining alive without evidence of local or distant tumor relapse at the time of analysis. OS represented the percentage of patients surviving at the time of censoring. LR was characterized by cancer recurrence limited to the pelvic region. The duration of follow-up spanned from the surgical intervention date to the final censoring point, with a minimum observational window of three years.

Statistical Analysis

All statistical analyses were carried out using IBM SPSS Statistics software (version 25.0). For variables exhibiting a normal distribution, results were expressed as the mean ± standard deviation and compared using Student’s t-test. Non-normally distributed data were summarized as medians (interquartile range) and evaluated using suitable nonparametric methods, such as the Mann-Whitney U test or the Kruskal-Wallis test, depending on the comparison group. Prognostic outcomes for patients categorized into low- and high-risk groups based on MRI characteristics were analyzed using the Kaplan-Meier survival estimation method, with differences between groups assessed via log-rank testing. To determine the prognostic relevance of specific variables for OS and DFS, both univariate and multivariate Cox proportional hazards regression models were employed. Results were reported as hazard ratios (HRs) with 95% confidence intervals (CIs). A two-sided p-value below 0.05 was considered statistically significant.

Results

Characteristics of Patients

A total of 302 individuals were included in the present study, with a median age of 56 years (interquartile range: 47–65). An overview of demographic characteristics and radiological staging data is provided in Table 1. Based on pre-treatment MRI features, 146 patients (48.4%) were categorized as high-risk, while the remaining 156 (51.6%) were identified as low-risk. In the patients administered NAT, 129 (42.7%) received the full dose of long-course chemoradiotherapy (LCCRT), which comprised of 113 (37.4%) treated with FOLFOX plus radiotherapy, 16 (5.3%) patients treated with XELOX plus radiotherapy. One (0.3%) patient was unable to complete the LCCRT, while another 3 (1%) patients received short-course radiotherapy. Additionally, 169 (56%) patients received only preoperative chemotherapy, which included 6 (2.0%) patients treated with XELOX, 163 (54.0%) patients treated with FOLFOX. 28 (9.3%) patients were treated by abdominoperineal excision, and 274 (90.7%) patients underwent restorative anterior resection (Table 2). The median interval from completion of NAT to surgery was 4 weeks (interquartile range, 3-7 weeks). The median interval was 4 weeks (interquartile range, 3-8 weeks) in the baseline MRI low-risk group and 4 weeks (interquartile range, 3-6 weeks) in the baseline MRI high-risk group, with no significant difference between groups (p = 0.284). Adjuvant chemotherapy was administered in 239 cases (79.1%), comprising 128 patients from the low-risk cohort and 111 from the high-risk cohort, with no statistically significant difference observed (p = 0.198). Post-NAT assessment using MRI reclassification showed that 115 patients (38.1%) remained in or shifted to the high-risk group, while 187 (61.9%) were categorized as low-risk. Notably, 35 of the 146 patients initially deemed MRI high-risk were reclassified as yMRI low-risk following NAT. All participants initially presented with clinical T3-4 or N+ status, and 122 individuals demonstrated radiologic downstaging of the TNM classification—either from T3–T4 to T1–T2 or from N1–N2 to N0—after completing NAT.

Table 1.

Characteristics of Patients

		MRI low-risk	MRI high-risk	P value
	Total 302	156(51.6%)	146(48.4%)
Age in years, median (IQR)	56(47,65)	55(48,65)	57(47,64)	0.801
BMI in Kg/m2, median (IQR)	22.2(20.6,24.5)	22.6(20.8,24.3)	22.1(20.5,24.5)	0.557
T stage diagnosed by MRI, n (%)				0.117
T1-T2	6(1.9)	5(3.2)	1(0.7)
T3-T4	296(98.1)	151(96.8)	145(99.3)
N stage diagnosed by MRI, n (%)				<0.001
N0	68(22.5)	41(26.3)	27(18.5)
N1-N2	234(77.5)	115(73.7)	119(81.5)
CRM status diagnosed by MRI, n (%)				NA
Clear	217(71.9)	156(100)	61(41.8)
Involved	85(28.1)	0(0)	85(58.2)
EMVI diagnosed by MRI, n (%)				NA
Negative	213(70.5)	156(100)	57(39.1)
Positive	89(29.5)	0(0)	89(60.9)
TD diagnosed by MRI, n (%)				NA
Negative	277(91.7)	156(100)	121(82.9)
Positive	25(8.3)	0(0)	25(17.1)
T stage diagnosed by yMRI, n (%)				0.002
T1-T2	34(11.3)	26(16.7)	8(5.5)
T3-T4	268(88.7)	130(83.3)	138(94.5)
N stage diagnosed by yMRI, n (%)				0.071
N0	163(53.9)	92(58.9)	71(48.6)
N1-2	139(46.1)	64(41.1)	75(51.4)
CRM status diagnosed by yMRI, n (%)				<0.001
Clear	236(78.1)	152(97.4)	84(57.5)
Involved	66(21.9)	4(2.6)	62(42.5)
EMVI diagnosed by yMRI, n (%)				<0.001
Negative	238(78.8)	154(98.7)	84(57.5)
Positive	64(21.2)	2(1.3)	62(42.5)
TD diagnosed by yMRI, n (%)				NA
Negative	284(94.1)	156(100)	128(87.7)
Positive	18(5.9)	0(0)	18(12.3)
Tumor height, n (%)				0.751
>5 cm	173(57.3)	88(56.4)	85(58.2)
<5 cm	129(42.7)	68(43.6)	61(41.8)
Adjuvant therapy, n(%)
Yes	239(79.1)	128(82.1)	111(76.1)	0.198
No	63(20.9)	28(17.9)	35(23.9)

Note. Data in the table presents the baseline characteristics of patients categorized into MRI low-risk and MRI high-risk groups.

Abbreviations: BMI: body mass index; MRI: magnetic resonance imaging; CRM: circumferential resection margin; EMVI: extramural vascular invasion; TD: tumor deposit.

Table 2.

Treatment of Patients

		MRI low-risk	MRI high-risk	P value
	Total 302	156(51.7%)	146(48.3%)
Neoadjuvant chemotherapy, n(%)
FOLFOX	279(92.4%)	145(92.9%)	134(91.8%)
XELOX	23(7.6%)	11(7.1%)	12(8.2%)	0.829
Radiotherapy, n(%)
Yes	133(44.0%)	72(46.2%)	61(41.8%)
No	169(56.0%)	84(53.8%)	85(58.2%)	0.446
Time interval from NAT to surgery, median (IQR)	4(3,7)	4(3,8)	4(4,6)	0.284
Surgery
Abdominoperineal excision	28(9.3%)	14(9.0%)	14(9.6%)
Restorative anterior resection	274(90.7%)	142(91.0%)	132(90.4%)	0.854
Adjuvant therapy, n(%)
Yes	239(79.1%)	128(82.1%)	111(76.0%)
No	63(20.9%)	28(17.9%)	35(24.0%)	0.198

Abbreviations: NAT: neoadjuvant therapy.

Prognosis According to MRI Features Before and After NAT

Among the 302 patients evaluated in this study, the median duration of postoperative follow-up was 30 months, ranging from 1 to 64 months. By the time of the final follow-up, 38 individuals (12.5%) had passed away, and 56 patients (18.5%) experienced disease recurrence. Among those with recurrence, 13 cases involved pulmonary metastases, 5 had hepatic metastases, another 5 presented with nonregional lymphatic metastases, 15 exhibited metastases to two or more distant organs, and 18 patients experienced locoregional recurrence. When stratified by initial MRI risk classification, the high-risk cohort showed significantly poorer oncologic outcomes, including lower three-year DFS (54.8% vs. 93.6%; p < 0.001), reduced three-year OS (76% vs. 98.1%; p < 0.001), and a higher three-year LR rate (10.9% vs. 1.3%; p = 0.002) compared to their low-risk counterparts. Notably, the prognostic significance of MRI-based risk stratification remained evident following NAT. Patients in the post-NAT yMRI high-risk category demonstrated inferior three-year DFS (47.8% vs. 91.4%; p < 0.001), lower three-year OS (71.3% vs. 97.3%; p < 0.001), and a markedly higher rate of LR over three years (13% vs. 1.6%; p < 0.001) in comparison to those classified as yMRI low-risk (Figure 2).

Figure 2.

Survival outcomes according to MRI features Disease-free survival (A), Overall survival (B) and Local recurrence (C) According to risk criteria based on MRI features at baseline. Disease-free survival (D), Overall survival (E) and Local recurrence (F) According to risk criteria based on MRI features after NAT

Table 3 summarizes the hazard ratios (HRs) for each prognostic variable in relation to survival outcome, as assessed through both univariate and multivariate analyses. Findings from both models indicated that CRM involvement, EMVI, and the presence of TDs, as detected on post-NAT MRI, were significantly associated with worse three-year DFS outcomes. Table 4 displays the univariate and multivariate HRs for each prognostic factor on three-year OS. In the univariate model, nodal stage N1–N2, yMRI-indicated CRM involvement, EMVI, and TDs were all linked to inferior three-year OS. Multivariate analysis further identified yMRI-based CRM involvement, the presence of TDs, and tumor height less than 5 cm as independent risk factors for poor three-year OS.

Table 3.

Cox Regression Analyses of the Prognostic Factors for Disease-free Survival Rate

	Univariable HR (95% CI)	P value	Multivariable HR (95% CI)	P value
Age in years
<65	1	0.797	1	0.561
>65	0.93(0.55-1.56)		0.85(0.49-1.45)
T stage diagnosed by yMRI
T1-T2	1	0.019	1	0.055
T3-T4	3.97(1.25-12.61)		3.16(0.97-10.26)
N stage diagnosed by yMRI
N0	1	0.058	1	0.321
N1-N2	1.54(0.98-2.41)		1.26(0.79-2.01)
CRM status diagnosed by yMRI
Clear	1	<0.001	1	<0.001
Involved	3.74(2.39-5.84)		2.71(1.66-4.41)
EMVI diagnosed by yMRI
Negative	1	<0.001	1	<0.001
Positive	3.81(2.43-5.97)		2.67(1.64-4.35)
TD diagnosed by yMRI
Negative	1	<0.001	1	<0.001
Positive	4.23(2.37-7.55)		4.31(2.36-7.88)
Tumor height
>5cm	1	0.545	1	0.136
<5cm	0.54(1.14-1.78)		1.41(0.89-2.23)

Note: Data in the table presents the univariable and multivariable Cox regression analyses of various prognostic factors for disease-free survival in rectal cancer patients. Bold values indicate statistical significance at P < 0.05.

Abbreviations: yMRI: magnetic resonance imaging after neoadjuvant therapy; CRM: circumferential resection margin; EMVI: extramural vascular invasion; TD: tumor deposit.

Table 4.

Cox Regression Analyses of the Prognostic Factors for Overall Survival Rate

	Univariable HR (95% CI)	P value	Multivariable HR (95% CI)	P value
Age in years
<65	1	0.223	1	0.551
>65	1.51(0.77-1.96)		1.23(0.61-2.48)
T stage diagnosed by yMRI
T1-T2	1	0.158	1	0.303
T3-T4	2.78(0.67-11.58)		2.16(0.49-9.43)
N stage diagnosed by yMRI
N0	1	0.048	1	0.253
N1-N2	1.93(1.01-3.69)		1.49(0.75-2.95)
CRM status diagnosed by yMRI
Clear	1	<0.001	1	<0.001
Involved	6.39(3.33-12.67)		5.12(2.48-10.55)
EMVI diagnosed by yMRI
Negative	1	<0.001	1	0.023
Positive	3.71(1.95-7.05)		2.25(1.12-4.54)
TD diagnosed by yMRI
Negative	1	<0.001	1	<0.001
Positive	4.31(1.89-9.77)		5.12(2.16-12.14)
Tumor height
>5cm	1	0.13	1	0.017
<5cm	1.64(0.86-3.13)		2.25(1.15-4.41)

Note: Data in the table presents the univariable and multivariable Cox regression analyses of various prognostic factors for overall survival in rectal cancer patients. Bold values indicate statistical significance at P < 0.05.

Abbreviations: yMRI: magnetic resonance imaging after neoadjuvant therapy; CRM: circumferential resection margin; EMVI: extramural vascular invasion; TD: tumor deposit.

Prognosis According to Response to NAT

Since MRI-detected CRM involvement, EMVI, and TDs before and after NAT were linked to worse survival, we investigated whether MRI-defined high-risk patients could still be further stratified after NAT. We analyzed 146 patients initially classified as MRI high-risk. Of these, 35 patients responded well to NAT and were defined as good responders. Their three-year DFS was 82.9%, OS was 94.3%, and LR was 2.9%. The other 111 patients were poor responders, showing a three-year DFS of 45.9%, OS of 70.3%, and LR of 13.5%. The differences between the two groups were statistically significant for DFS (p < 0.001), OS (p = 0.003), and LR (p = 0.05) (Figure 3). These results suggest that patients who converted from MRI high-risk to low-risk after NAT had much better survival and lower recurrence than poor responders. Their outcomes were also similar to those who were MRI low-risk from the beginning (Fig. S1).

Figure 3.

Survival outcomes according to response to NAT Disease-free survival (A), Overall survival (B) and Local recurrence (C) According to response to NAT

To determine whether MRI-based response assessment might overlook good responders who showed TNM downstaging, we conducted further analysis by dividing patients into four subgroups. Among the 111 poor responders, no significant differences were found in three-year DFS (p = 0.123), OS (p = 0.188), or LR (p = 0.991) between those who had TNM downstaging and those who did not. In the group of 85 patients without TNM downstaging, 19 were classified as good responders based on MRI. These patients had favorable outcomes, with a three-year DFS of 84.2%, OS of 89.5%, and no local recurrence. In contrast, the 66 poor responders in the same group had significantly worse outcomes, with a three-year DFS of 43.9%, OS of 66.7%, and LR of 12.1%. The differences between good and poor responders in this subgroup were statistically significant for DFS (p = 0.004) and OS (p = 0.047), with a trend toward significance for LR (p = 0.062) (Figure 4).

Figure 4.

Survival outcomes according to change of MRI features and TNM stage Disease-free survival (A), Overall survival (B) and Local recurrence (C) According to response to NAT in patients without TNM downstaging. Disease-free survival (D), Overall survival (E) and Local recurrence (F) According to response to NAT in patients with poor response

Moreover, we performed additional analyses within the baseline MRI high-risk subgroup. In multivariable analysis, baseline mrEMVI and mrTD remained associated with worse DFS (p = 0.021 and p = 0.001, respectively), whereas baseline mrMRF was not (p = 0.493). On post-treatment MRI, ymrMRF, ymrEMVI, and ymrTD were all adverse factors for DFS (p = 0.036, p = 0.013, and p = 0.001, respectively). When MRI dynamic response and TNM downstaging were entered into the same multivariable model, persistent post-treatment MRI high-risk status remained associated with worse DFS (p < 0.001), whereas TNM downstaging was not independently associated with DFS (p = 0.497) (Supplementary Tables S1). In addition, among patients who were positive for a given feature at baseline, we compared those in whom the corresponding feature cleared after NAT with those with persistent positivity. Clearance of mrMRF was associated with more favorable DFS and OS, whereas the corresponding associations for mrEMVI and mrTD did not reach statistical significance (Supplementary Tables S2 and S3). Residual post-treatment MRI high-risk burden also provided further prognostic stratification within the baseline MRI high-risk subgroup.

Discussion

Identifying which rectal cancer patients are most likely to benefit from NAT remains a key clinical challenge. In our study, we used MRI features to classify patients into high- and low-risk groups before treatment. As expected, the high-risk group had significantly worse survival than the low-risk group, and this difference remained evident after NAT. However, among those initially identified as high-risk, patients who converted to low-risk status after NAT had survival outcomes similar to those who were low-risk from the beginning. This suggests that some high-risk patients can benefit from NAT if they respond well to treatment.

Numerous investigations have repeatedly confirmed that MRI-detected CRM involvement, EMVI, and TDs are established predictors of local and distant recurrence, as well as poor survival, with these associations persisting after NAT.^11,24-27 In our cohort, post-NAT MRI (yMRI) showed positive rates of 21.6% for CRM involvement, 21.6% for EMVI, and 5.8% for TDs. On multivariable regression analysis, each remained an independent predictor of inferior survival, regardless of yT or yN stage. Patients with favorable MRI-assessed responses to NAT achieved better outcomes. These findings highlight the value of MRI features for risk stratification and guiding NAT use in rectal cancer. In the present cohort, post-treatment MRI high-risk features remained informative after NAT, supporting the view that restaging MRI contributes prognostic information beyond baseline staging alone. Rather than simply separating baseline high- and low-risk groups, our results further show that prognosis within the baseline MRI high-risk population remains heterogeneous after NAT.

This heterogeneity became more evident in the additional subgroup analyses. Within the baseline MRI high-risk group, baseline mrEMVI and mrTD, as well as post-treatment ymrMRF, ymrEMVI, and ymrTD, were associated with worse DFS in baseline MRI high-risk patients. Persistent post-treatment MRI high-risk status remained independently associated with worse DFS even when TNM downstaging was considered in the same model, suggesting that MRI dynamic response may capture prognostic information beyond conventional radiologic downstaging. Further analyses showed that prognosis differed according to each individual MRI high-risk feature and that patients in whom mrMRF disappeared after NAT had more favorable DFS and OS than those with persistent positivity, whereas the corresponding comparisons for EMVI and TD were directionally similar but limited by sample size. These findings support the role of MRI high-risk features as practical imaging markers for risk-adapted management.

In the context of personalized medicine, where treatment strategies are tailored according to individual risk of local and distant recurrence, imaging plays a vital role in guiding preoperative decisions. Accurate imaging helps optimize treatment plans and reduces the risk of both overtreatment and undertreatment.²⁸ Besides, Aliyev et al reported that robotic intersphincteric resection for low rectal cancer was associated with better mesorectal integrity, no conversion to open surgery, and fewer postoperative complications than the laparoscopic approach, although long-term oncologic and anorectal functional outcomes were similar,²⁹ whereas Shadmanov et al showed in a 23-year retrospective series that the later treatment era, characterized by robotic surgery, high-resolution 3-T MRI, and updated management strategies, was associated with better DFS and OS together with a trend toward lower local recurrence and distant metastasis.³⁰ Taken together, these findings suggest that rectal cancer outcomes are shaped not only by tumor biology, but also by evolving imaging quality, surgical platforms, and treatment pathways. For this reason, routine NAT in low-risk patients offers limited survival benefit and can expose them to significant short-term toxicities and long-term side effects.³¹ Additionally, NAT-induced fibrosis in tumors and surrounding tissues can complicate TME, potentially compromising surgical outcomes.^32,33 Preoperative radiotherapy has also been linked to adverse effects, including postoperative pelvic floor dysfunction and sexual dysfunction.^34,35 Therefore, developing reliable methods to identify patients who are most likely to benefit from NAT is essential for improving outcomes and minimizing harm.

Recent advances in radiomics and artificial intelligence (AI) have opened new avenues for improving imaging-based risk stratification in rectal cancer. By extracting high-dimensional quantitative features from MRI, radiomic analysis can capture subtle spatial and textural patterns that are often imperceptible to the human eye. These features, when integrated with clinical and pathological parameters, have shown promise in predicting treatment response, tumor regression grade, and long-term outcomes. Furthermore, AI-based deep learning models can automatically learn hierarchical representations from imaging data, enabling more objective and reproducible assessments of tumor biology. In the future, combining conventional MRI risk factors such as CRM, EMVI, and TDs with AI-derived imaging biomarkers may allow for dynamic and individualized prediction of NAT efficacy, thereby enhancing decision-making within multidisciplinary teams. Integrating such data-driven tools into clinical workflows could substantially advance precision medicine in rectal cancer management.

This study has several limitations. First, as a retrospective single-center analysis, potential selection bias may have occurred, particularly due to the exclusion of patients with incomplete clinical or imaging records, which may favor more standardized cases. Second, pathologic validation of MRI-assessed EMVI and TDs was not performed, and thus minor discrepancies between imaging and histopathology cannot be excluded. Third, MRI technology and interpretation standards have evolved since the study period (2017–2019), which may affect the applicability of our findings to current clinical practice. Finally, the median follow-up duration of 30 months was relatively short, which may be insufficient for a comprehensive assessment of long-term outcomes such as 5-year DFS, OS, and local recurrence rates. The current results mainly reflect short-to mid-term findings, and longer-term prognostic differences may become more apparent with extended observation. Further studies with larger sample sizes and longer follow-up periods are warranted to validate the robustness and long-term clinical significance of our findings.

Conclusion

Baseline MRI high-risk features were associated with poor prognosis in rectal cancer. Among baseline MRI high-risk patients, MRI risk conversion after NAT may identify a subgroup with more favorable outcomes, whereas persistent post-treatment MRI high-risk features were associated with worse survival.

Supplemental Material

Supplemental material - Magnetic Resonance Imaging Features as Predictors of Response and Prognosis Following Neoadjuvant Therapy in Rectal Cancer: A Retrospective Cohort Study

Supplemental material for Magnetic Resonance Imaging Features as Predictors of Response and Prognosis Following Neoadjuvant Therapy in Rectal Cancer: A Retrospective Cohort Study by Jinhua Deng, Sijing Cheng, Shujuan Li, and Yujie Hou in Clinical Medicine Insights: Oncology.

Supplemental Material

Supplemental material - Magnetic Resonance Imaging Features as Predictors of Response and Prognosis Following Neoadjuvant Therapy in Rectal Cancer: A Retrospective Cohort Study

Footnotes

Acknowledgements

This study was supported by grants from the National Natural Science Foundation of China (82400599), Guangdong Provincial Clinical Research Center for Digestive Diseases (2020B1111170004), Innovative Clinical Technique of Guangzhou, and National Key Clinical Discipline.

ORCID iDs

Jinhua Deng

Yujie Hou

Ethical Considerations

This study was approved by the Institutional Review Board of The Sixth Affiliated Hospital, Sun Yat-sen University (approval number: 2022ZSLYEC-615; approval date: December 16, 2022) and conducted in accordance with the ethical standards of the institutional research committee and the Helsinki Declaration of 1975, as revised in 2024.

Consent for Publication

The requirement for informed consent was waived by the approving IRB due to the retrospective study design and full anonymization of all patient clinical data.

Author Contributions

Jinhua Deng: conceptualization, data curation, investigation, methodology, and writing-original draft. Sijing Cheng: formal analysis, methodology, visualization, and writing-review & editing. Shujuan Li: validation, formal analysis, and writing-review & editing. Yujie Hou: conceptualization, supervision, project administration, funding acquisition, and writing-review & editing. These authors contributed equally.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by grants from the National Natural Science Foundation of China (82400599), Guangdong Provincial Clinical Research Center for Digestive Diseases (2020B1111170004), Innovative Clinical Technique of Guangzhou, and National Key Clinical Discipline.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The original research data supporting the results, tables, and figures presented in this manuscript have been retained by the authors and can be made available by the corresponding author upon reasonable request, subject to institutional and ethical requirements regarding patient confidentiality.*

Use of Artificial Intelligence

No artificial intelligence tools were used in the writing, editing, analysis, or preparation of this manuscript.

Supplemental Material

Supplemental material for this article is available online.

References

Saraf

Roberts

Parikh

. Optimal neoadjuvant strategies for locally advanced rectal cancer by risk assessment and tumor location. J Natl Compr Canc Netw. 2022;20(10):1177-1184.

Scott

Kennedy

Berlin

, et al. Management of locally advanced rectal cancer: ASCO guideline. J Clin Oncol. 2024;42(28):3355-3375.

Benson

Venook

Adam

, et al. NCCN Guidelines Insights: rectal cancer, version 3.2024. J Natl Compr Canc Netw. 2024;22(6):366-375.

Bromham

Kallioinen

Hoskin

Davies

Guideline Committee . Colorectal cancer: summary of NICE guidance. BMJ. 2020;368:m461.

Glynne-Jones

Wyrwicz

Tiret

, et al. Rectal cancer: ESMO clinical practice guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2017;28(suppl 4):iv22-iv40.

van Gijn

Marijnen

Nagtegaal

, et al. Preoperative radiotherapy combined with total mesorectal excision for resectable rectal cancer: 12-year follow-up of the multicentre, randomised controlled TME trial. Lancet Oncol. 2011;12(6):575-582.

Cho

Choi

Kim

, et al. Long-term outcomes of surgery alone versus surgery following preoperative chemoradiotherapy for early T3 rectal cancer: a propensity score analysis. Medicine (Baltimore). 2017;96(12):e6362.

Ansari

Solomon

Fisher

, et al. Acute adverse events and postoperative complications in a randomized trial of preoperative short-course radiotherapy versus long-course chemoradiotherapy for T3 adenocarcinoma of the rectum: Trans-Tasman Radiation Oncology Group Trial (TROG 01.04). Ann Surg. 2017;265(5):882-888.

Lin

Zhang

Chi

, et al. Neoadjuvant short-course radiotherapy followed by camrelizumab and chemotherapy in locally advanced rectal cancer (UNION): early outcomes of a multicenter randomized phase III trial. Ann Oncol. 2024;35(10):882-891.

10.

Liu

Zeng

Zhang

, et al. Morbidity, mortality, and pathologic outcomes of transanal versus laparoscopic total mesorectal excision for rectal cancer: short-term outcomes from a multicenter randomized controlled trial. Ann Surg. 2023;277(1):1-6.

11.

Agger

Jörgren

Lydrup

Buchwald

. Risk of local recurrence of rectal cancer and circumferential resection margin: population-based cohort study. Br J Surg. 2020;107(5):580-585.

12.

Lord

D'Souza

Shaw

, et al. MRI-diagnosed tumor deposits and EMVI status have superior prognostic accuracy to current clinical TNM staging in rectal cancer. Ann Surg. 2022;276(2):334-344.

13.

Mazaki

Tsukamoto

Miyake

, et al. Circumferential resejction margin status as a predictive factor for recurrence in preoperative MRI for advanced lower rectal cancer without preoperative therapy. Dis Colon Rectum. 2021;64(1):71-80.

14.

Lord

Corr

Chandramohan

, et al. Assessment of the 2020 NICE criteria for preoperative radiotherapy in patients with rectal cancer treated by surgery alone in comparison with proven MRI prognostic factors: a retrospective cohort study. Lancet Oncol. 2022;23(6):793-801.

15.

von Elm

Altman

Egger

, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol. 2008;61(4):344-349.

16.

Taylor

Quirke

Heald

, et al. Preoperative magnetic resonance imaging assessment of circumferential resection margin predicts disease-free survival and local recurrence: 5-year follow-up results of the MERCURY study. J Clin Oncol. 2014;32(1):34-43.

17.

Battersby

How

Moran

, et al. Prospective validation of a low rectal cancer magnetic resonance imaging staging system and development of a local recurrence risk stratification model: the MERCURY II study. Ann Surg. 2016;263(4):751-760.

18.

Chen

Tang

, et al. The prognostic value of MRI-detected extramural vascular invasion (mrEMVI) for rectal cancer patients treated with neoadjuvant therapy: a meta-analysis. Eur Radiol. 2021;31(12):8827-8837.

19.

Inoue

Sheedy

Heiken

, et al. MRI-detected extramural venous invasion of rectal cancer: multimodality performance and implications at baseline imaging and after neoadjuvant therapy. Insights Imaging. 2021;12(1):110.

20.

Amin

Greene

Edge

, et al. The eighth edition AJCC cancer staging manual: continuing to build a bridge from a population-based to a more “personalized” approach to cancer staging. CA Cancer J Clin. 2017;67(2):93-99.

21.

Cianci

Cristel

Agostini

, et al. MRI for rectal cancer primary staging and restaging after neoadjuvant chemoradiation therapy: how to do it during daily clinical practice. Eur J Radiol. 2020;131:109238.

22.

Horvat

Veeraraghavan

Khan

, et al. MR imaging of rectal cancer: radiomics analysis to assess treatment response after neoadjuvant therapy. Radiology. 2018;287(3):833-843.

23.

Patel

Brown

Machado

, et al. MRI assessment and outcomes in patients receiving neoadjuvant chemotherapy only for primary rectal cancer: long-term results from the GEMCAD 0801 trial. Ann Oncol. 2017;28(2):344-353.

24.

Agger

Jörgren

Jöud

Lydrup

Buchwald

. Negative prognostic impact of tumor deposits in rectal cancer: a national study cohort. Ann Surg. 2023;278(3):e526-e533.

25.

Kim

Firat

Thompson

, et al. Extramural venous invasion and tumor deposit at diffusion-weighted MRI in patients after neoadjuvant treatment for rectal cancer. Radiology. 2023;308(2):e230079.

26.

Nougaret

Gormly

Lambregts

DMJ

, et al. MRI of the rectum: a decade into DISTANCE, moving to DISTANCED. Radiology. 2025;314(1):e232838.

27.

Rouleau Fournier

Motamedi

MAK

Brown

, et al. Oncologic outcomes associated with MRI-detected extramural venous invasion (mrEMVI) in rectal cancer: a systematic review and meta-analysis. Ann Surg. 2022;275(2):303-314.

28.

Guan

Zhang

, et al. Contrast-enhanced MRI for T restaging of locally advanced rectal cancer following neoadjuvant chemotherapy and radiation therapy. Radiology. 2022;305(2):364-372.

29.

Aliyev

Piozzi

Bulut

, et al. Robotic vs. laparoscopic intersphincteric resection for low rectal cancer: a case matched study reporting a median of 7-year long-term oncological and functional outcomes. Updates Surg. 2022;74(6):1851-1860.

30.

Shadmanov

Aliyev

Bakır

Goksel

Asoglu

. The shifting and evolving neoadjuvant treatments and surgical platforms on oncological outcomes and sphincter preservation in distal rectal cancer: a 23-year retrospective experience. J Gastrointest Cancer. 2025;56(1):177.

31.

Wang

Jia

Gao

, et al. Serum metabolic traits reveal therapeutic toxicities and responses of neoadjuvant chemoradiotherapy in patients with rectal cancer. Nat Commun. 2022;13(1):7802.

32.

Jia

Wang

Cheng

, et al. Low-versus high-risk rectal cancer based on MRI features: outcomes in patients treated without neoadjuvant chemoradiotherapy. AJR Am J Roentgenol. 2018;211(2):327-334.

33.

Zeng

Luo

Zhang

, et al. Transanal vs laparoscopic total mesorectal excision and 3-year disease-free survival in rectal cancer: the TaLaR randomized clinical trial. JAMA. 2025;333(9):774-783.

34.

Marshall

Tarras

Ali

Bloom

Torres

Kahn

. Female erectile tissues and sexual dysfunction after pelvic radiotherapy: a scoping review. CA Cancer J Clin. 2022;72(4):353-359.

35.

Qin

Huang

, et al. Development and validation of a post-radiotherapy prediction model for bowel dysfunction after rectal cancer resection. Gastroenterology. 2023;165(6):1430-1442.e1414.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

1.20 MB

0.01 MB

0.00 MB