Association of Pelvic Structure Involvement and Tumor Morphology at MRI with Prognosis Following Resection in Locally Recurrent Rectal Cancer

Abstract

Purpose

To determine the influence of location, extent of tissue invasion, and tumor morphology at MRI on the resectability of locally recurrent rectal cancer (LRRC) and postresection oncologic outcomes of LRRC.

Materials and Methods

This retrospective observational study included consecutive patients diagnosed with LRRC who underwent surgery with curative intent at the Catharina Hospital Eindhoven and Karolinska University Hospital Stockholm between January 2003 and December 2017. Two expert radiologists reviewed available MR images while adhering to a standardized reviewing checklist. The effect of pelvic structure involvement, tumor morphology on the primary outcome of resection margin status, and secondary outcomes of overall survival and disease-free survival were assessed using univariable and multivariable logistic regression and Cox proportional hazard analyses.

Results

The final analysis included 328 patients with LRRC (mean age ± SD, 64.9 years ± 9.6; 126 female, 202 male). Resection margins were negative in 217 (66.2%) patients and positive in 111 patients (33.8%). Tumor size, tumor type, and border type on MR images were all associated with resectability. Central recurrences were associated with the lowest likelihood of positive resection margins (odds ratio [OR], 0.45; 95% CI: 0.28, 0.71; P < .001), whereas lateral recurrences were associated with the highest likelihood (OR, 2.00; 95% CI: 1.25, 3.19: P = .004). Similarly, central recurrences were associated with better disease-free survival compared with lateral recurrences (hazard ratio [HR], 0.69; 95% CI: 0.53, 0.90; P = .006 vs HR, 1.49; 95% CI: 1.14, 1.94; P = .003, respectively). Similar findings were observed after correcting for resection margin status.

Conclusion

Standardized MRI assessment of tumor characteristics in patients with LRRC resulted in the identification of specific prognostic factors. Central compartment involvement and well-defined tumors were associated with improved prognosis, whereas lateral compartment involvement and fibrotic spiculated tumors were associated with a worse prognosis after surgical resection.

Keywords: Rectum, MR-Imaging, Abdomen/GI, Oncology, Surgery, Locally Recurrent Rectal Cancer, Tumor Biology

Supplemental material is available for this article.

© RSNA, 2025

Summary

Standardized MRI assessment of tumor characteristics in patients with locally recurrent rectal cancer identified central compartment involvement and well-defined tumors as characteristics associated with improved prognosis, whereas lateral compartment involvement and fibrotic spiculated tumors were associated with a worse prognosis following surgical resection.

Key Points

• ■ In this retrospective study of 328 patients with locally recurrent rectal cancer, a standardized reporting checklist was used to identify specific MRI-derived tumor characteristics.

• ■ Central compartment involvement (hazard ratio, 0.69; 95% CI: 0.53, 0.90; P = .006) and well-defined tumors (odds ratio, 0.37; 95% CI: 0.19, 0.71; P = .003) were significant factors associated with an improved prognosis after surgical resection.

• ■ Lateral compartment involvement (hazard ratio, 1.49; 95% CI: 1.14, 1.94; P = .003) and fibrotic spiculated tumors (hazard ratio, 1.63; 95% CI: 1.02, 2.62; P = .04) were associated with worse prognosis after surgical resection.

Introduction

In locally recurrent rectal cancer (LRRC), negative resection margins (ie, R0 resection) are strongly associated with better overall survival (OS) and disease-free survival (DFS) (1,2). Successfully achieving an R0 resection can be challenging and may require extensive surgery, which is associated with high, although acceptable, morbidity (3–5). It is of utmost importance for clinicians to appropriately select patients in whom surgical treatment could be beneficial (2). Moreover, for patients with poor tumor biology, it is unlikely that resection margin status will substantially impact survival. Identifying patients in whom extensive surgery will not be of additional prognostic value is one of the most important goals of multidisciplinary discussions involved in the treatment of LRRC.

Previous research has shown that centrally located LRRC has the most favorable prognosis, followed by recurrences in the anterior or posterior compartment (6–11). Involvement of the lateral compartment is associated with the poorest survival (8,10,12). Current literature attributes these varying prognoses mainly to the probability of achieving an R0 resection, which is influenced by the location of the recurrence and the extent of invasion into surrounding structures (8,12,13). However, the prognosis of LRRC located in different compartments is also presumably affected by both the etiology and the biologic behavior of the recurrence (14,15). For instance, lateral recurrences have a higher chance of death after treatment when adjusted for age, sex, and resection margin status (12). This suggests more aggressive biologic behavior of lateral LRRC, independently of the apparent success of surgical treatment.

In primary rectal cancer, MRI-detected extramural venous invasion, tumor deposits, tumor type (solid, fibrotic, or mucinous), and tumor border type (well-defined or spiculated or irregular) are associated with survival-related outcomes (16–20). Because these characteristics are considered as early indicators of tumor biologic behavior in primary rectal cancer, they are routinely reported in current radiologic reports. In LRRC, the influence of these and other radiologic characteristics on tumor resectability and oncologic outcomes is less well reported and thus not regularly included in radiologic reports (21,22).

The primary aim of the current study was to determine the influence of location, extent of tissue invasion, and tumor morphology at MRI on the resectability of LRRC and postresection oncologic outcomes of LRRC. A standardized MRI reporting checklist for LRRC was used for interpretation of tumor features.

Materials and Methods

Patients and Treatment

This study was reviewed and approved by the Dutch Medical Research Ethics Committees United (registration no. W23.008) and Swedish Ethical Review Authority (registration no. 2015/1547 and 2024-01244-02). The study was not subject to the Medical Research Involving Human Subjects Act. Written informed consent for participation was waived due to the retrospective nature of this study. All patients with LRRC without synchronous metastases who underwent surgical treatment with curative intent between January 2003 and December 2017 at the Catharina Hospital Eindhoven or Karolinska University Hospital Stockholm were included. Patients were ineligible if no data or radiologic imaging on tumor involvement or invasion into surrounding pelvic structures were available, when no rectal resection for the primary tumor was performed, or if they presented with a local recurrence or beyond. The oncologic outcomes of our study sample (n = 328) were previously reported by Nordkamp et al (23). However, the current study is unique in that the oncologic outcomes were analyzed in relation to specific MRI-derived tumor characteristics.

All patient cases were reviewed in a multidisciplinary discussion before and during treatment at both institutions, with the multidisciplinary team consisting of at least one surgical oncologist, one medical oncologist, one radiologist, one radiation oncologist, and one pathologist. At Catharina Hospital Eindhoven, all patients received neoadjuvant treatment. Based on case-specific characteristics, the multidisciplinary team assessed whether patients should receive induction chemotherapy. Patients previously treated with irradiation underwent reirradiation (30 Gy), whereas those without a history of pelvic irradiation received full-course irradiation (50–50.4 Gy), both in conjunction with oral capecitabine. Patients classified as borderline resectable or unresectable received induction chemotherapy, followed by irradiation or reirradiation alongside oral capecitabine, aiming for maximal downstaging. At Karolinska University Hospital Stockholm, neoadjuvant treatment was limited to radiation therapy–naive patients according to the local routine at the time. Neither induction chemotherapy nor reirradiation was used. However, most of the patients not previously treated with irradiation received neoadjuvant full course chemoradiation therapy and/or chemotherapy.

At the multidisciplinary discussion, the type of surgery was determined depending on the location of the tumor and/or the involvement of surrounding structures. All performed procedures were subdivided into the following groups: low anterior resection, abdominoperineal excision, multivisceral resection, posterior exenteration, total exenteration, extra-anatomic resection, and hemipelvectomy.

Data Collection

Data were collected retrospectively. Retrieved data consisted of patient and tumor characteristics, radiologic imaging, neoadjuvant treatment, surgical procedures, pathologic data, and follow-up data. During follow-up, the occurrence of local recurrence or distant metastases was recorded. Survival data were obtained by the Municipal Administrative Databases. Follow-up data were obtained from the medical records.

Image Acquisition and Analysis

All patients underwent a baseline pelvic MRI, consisting of T2-weighted images in two orthogonal planes, preferably transversal and sagittal, at a minimum section thickness of 5 mm. From 2008 onward, diffusion-weighted imaging was deemed standard, and from 2015 onward, minimum section thickness was set at 3 mm. Pelvic MRI was repeated after neoadjuvant treatment, if applicable. Thoracoabdominal CT or fluorodeoxyglucose PET/CT scans were performed to evaluate the presence of distant metastases.

For this study, pelvic MRI scans were reassessed by two expert radiologists in consensus, one at Catharina Hospital Eindhoven (J.N.) and one at Karolinska University Hospital Stockholm (C.S.). Both readers are abdominal radiologists with multiple years (J.N., with 9 years and C.S., with 15 years) of clinical and scientific experience. All cases were read by both readers; in case of disagreement, consensus was achieved between both readers during online meetings. If the quality of the pelvic MRI scans was deemed sufficient, reassessment was carried out according to a standardized reporting checklist (Appendix S1), ensuring a standardized and similar assessment of the MRI scans. During reassessment of the pelvic MRI scans, the following variables were reported: number of lesions; tumor size; tumor type (ie, mucinous, solid, fibrotic); border type (well-defined or spiculated or irregular); primary involved compartment; involved compartments; and involved pelvic structures per compartment. The primary end point was resection margin status. Secondary end points included OS and DFS.

Statistical Analysis

Statistical analyses were performed using SPSS Statistics version 29.0 software (IBM). OS was calculated from the date of surgery until the date of death, or the patient was censored at the last date of follow-up. DFS was calculated from the date of surgery until the date of diagnosis of local recurrence or metastases or death. Patients were censored if the relevant outcome did not occur during the study period due to loss to follow-up or the end of the study period. Continuous data are reported as means with SDs or medians with IQRs, depending on parameter distribution. Categorical data are reported as counts with percentages. For categorical data, group comparisons were performed using χ² test or Fisher exact test, as appropriate. Survival rates and cumulative incidences were calculated using Kaplan-Meier analysis. Comparisons were made using log-rank test. Odds ratios (ORs) were calculated to assess the association of pelvic compartment involvement, pelvic structure involvement, and tumor morphology with resection margin status, using both univariable and multivariable logistic regression analyses. Hazard ratios (HRs) were calculated to assess the association of pelvic compartment involvement and tumor morphology with OS and DFS using Cox proportional hazard analyses. Comparisons were made between patients with involvement of specific compartments or anatomic structures and those without such involvement (eg, all patients with central compartment involvement were compared with those without). No corrections were applied to the significance threshold due to the exploratory nature of the study. Statistical significance for all tests or comparisons was defined as a P value less than .05. In univariable analyses, statistical significance was defined as a P value less than .10. Univariable factors below this threshold were entered into multivariable regression analyses. In case of significant collinearity, variables were omitted. R0 resection was defined as the distance of the resection margins to the nearest tumor cells being more than 0 mm, R1 resection was defined as this distance being less than or equal to 0 mm. Corrections for resection margin status, if applicable, were performed by solely selecting R0 cases, thereby eliminating the effect of positive resection margins on survival.

Results

Patients

A total of 377 patients were treated for LRRC between January 2003 and December 2017. The final analysis included 328 patients (mean age ± SD, 64.9 years ± 9.6; 126 female, 202 male) with sufficient data quality to determine the extent of involvement and tumor morphology (Fig 1). Of these 328 patients, 219 patients were treated at the Catharina Hospital Eindhoven and 109 patients at the Karolinska University Hospital Stockholm. Table 1 shows baseline patient, primary tumor, local recurrence, and treatment characteristics and the distribution of involved compartments as determined using MRI. Median follow-up was 36 months (IQR, 18–61 months).

Figure 1:

Flowchart of patient selection. LRRC = locally recurrent rectal cancer.

Table 1:

Baseline Clinical and MRI-derived Characteristics of Patients with Locally Recurrent Rectal Cancer

Characteristic	Value
Patient (n = 328)
Sex
Female	126 (38.4)
Male	202 (61.6)
Age at surgery (y)	64.9 ± 9.6
Hospital
Catharina Hospital Eindhoven	219 (66.8)
Karolinska University Hospital Stockholm	109 (33.2)
Primary tumor
Neoadjuvant (chemo)radiation therapy
None	109 (33.2)
Radiation therapy alone	128 (39.0)
Chemoradiotherapy	91 (27.7)
Surgery primary tumor
Recto-sigmoidal resection	51 (15.5)
Low anterior resection	163 (49.7)
Abdominoperineal resection	108 (32.9)
Unknown	6 (1.8)
pT stage primary tumor
0–2	64 (19.5)
3	212 (64.6)
4	38 (11.6)
Unknown	14 (4.3)
pN stage primary tumor
0	141 (43.0)
1 or 2	172 (52.4)
Unknown	15 (4.6)
Adjuvant chemotherapy	69 (21.0)
Local recurrence
No. of lesions
Unifocal	302 (92.1)
Multifocal	26 (7.9)
Tumor type
Solid	245 (74.7)
Mucinous	58 (17.7)
Fibrotic	24 (7.3)
Tumor border type
Well-defined	71 (21.6)
Spiculated or irregular	257 (78.4)
Compartment involvement
Central compartment	196 (59.8)
Anterior compartment	209 (63.7)
Posterior compartment	145 (44.2)
Lateral compartment	170 (51.8)
Treatment
Induction chemotherapy	63 (19.2)
Neoadjuvant treatment
(Chemo)radiation therapy (45–50 Gy)	85 (25.9)
(Chemo)reirradiation (30 Gy)	151 (46.0)
LRRC surgery
Resection with reanastomosis	23 (7.0)
Abdominoperineal resection	51 (15.5)
Multivisceral resection	160 (48.8)
Extra-anatomic resection	27 (8.2)
Total exenteration	63 (19.2)
Hemipelvectomy	4 (1.2)
Intraoperative radiation therapy	202 (61.6)
Resection margin
R0	217 (66.2)
R1 or R2	111 (33.8)

Note.—Data are frequencies with percentages in parentheses or means ± SDs. LRRC = locally recurrent rectal cancer, R0 = radical resection, R1/R2 = irradical resection.

Association of Compartment Involvement with Resection Margin Status

Single compartment involvement was observed in 22.9% (75 of 328) of patients, and 77.1% (253 of 328) had involvement of two or more compartments. Central compartment involvement was observed at MRI in 59.8% (196 of 328) of patients, anterior compartment involvement in 63.7% (209 of 328), posterior compartment involvement in 44.2% (145 of 328), and lateral compartment involvement in 51.8% (170 of 328) (Fig 2).

Figure 2:

T2-weighted MR images of primary involved compartments in locally recurrent rectal cancer. (A) Non–contrast-enhanced T2-weighted sagittal MR image shows a solid tumor type in a 58-year-old male patient. (B) Non–contrast-enhanced T2-weighted transverse MR image of a mucinous tumor type in a 48-year-old male patient. (C) Non–contrast-enhanced T2-weighted transverse MR image of a fibrotic tumor type in a 75-year-old-male patient. Tumors are circled in blue.

The number of resections with positive margins (R1 resection) for tumors with central compartment involvement was significantly lower compared with tumors involving all other compartments (26.5% [52 of 196] vs 44.7% [59 of 132]; P < .001; Table 2), and central compartment involvement was associated with lower risk of R1 resection at univariable analysis (OR, 0.45; 95% CI: 0.28, 0.71; P < .001; Table 3). Overall, anterior compartment involvement did not influence the likelihood of an R1 resection. The number of R1 resections was significantly higher in patients with posterior and lateral compartment involvement compared with all other compartments (40.7% [59 of 145] vs 28.4% [52 of 183]; P = .01 and 41.2% [70 of 170] vs 25.9% [41 of 158]; P = .003, respectively). Posterior and lateral compartment involvement both significantly increased the likelihood of an R1 resection (OR, 1.73; 95% CI: 1.09, 2.74; P = .02 and OR, 2.00; 95% CI: 1.25, 3.19; P = .004, respectively).

Table 2:

R0 and R1 Resection Rate According to Involved Compartment and Tumor Characteristics at Pretreatment MRI

Parameter	Negative Resection Margins (R0)	Positive Resection Margins (R1)	Total	P Value
Central compartment				<.001*
Involved	144 (73.5)	52 (26.5)	196
Not involved	73 (55.3)	59 (44.7)	132
Anterior compartment				.12
Involved	133 (63.6)	76 (36.4)	209
Not involved	84 (78.7)	35 (29.4)	119
Posterior compartment				.01*
Involved	86 (59.3)	59 (40.7)	145
Not involved	131 (71.6)	52 (28.4)	183
Lateral compartment				.003*
Involved	100 (58.8)	70 (41.2)	170
Not involved	117 (74.1)	41 (25.9)	158
No. of lesions				.84
1	201 (66.6)	101 (33.4)	302
2	12 (63.2)	7 (36.8)	19
3	4 (57.1)	3 (42.9)	7
Tumor type				.03*
Mucinous	47 (81.0)	11 (19.0)	58
Solid	154 (62.9)	91 (37.1)	245
Fibrotic	16 (66.7)	8 (33.3)	24
Other	0	1 (100)	1
Tumor border type				.001*
Well-defined	58 (81.7)	13 (18.3)	71
Spiculated or irregular	159 (62.1)	97 (37.9)	256
Unknown	0	1 (100)	1

Note.—Data are numbers with percentages in parentheses. R0 = radical resection, R1 = irradical resection.

^*Statistically significant.

Table 3:

Association of Pelvic Structure Involvement to Surgical Resection Margins

Predictors	Univariable		Multivariable
	Odds Ratio (95% CI)	P Value	Odds Ratio (95% CI)	P Value
Central structure involvement	0.45 (0.28, 0.71)	<.001*	—
Anastomotic	0.52 (0.32, 0.84)	.008*	—
Levator muscle	0.80 (0.40, 1.59)	.52	—
Infralevator perineum	0.69 (0.30, 1.59)	.38	—
Gluteus muscle	0.64 (0.13, 3.24)	.59	—
Anterior structure involvement	1.37 (0.85, 2.23)	.20	—
Vagina	0.53 (0.21, 1.35)	.18	—
Uterus	0.56 (0.17, 1.82)	.34	—
Prostate	2.01 (1.06, 3.80)	.03*	1.00 (0.99, 1.00)	.17
Ureter	2.43 (1.43, 4.13)	.001*	2.42 (1.42, 4.12)	.001*
Urinary bladder	1.14 (0.63, 2.05)	.68	—
Urethra	0.78 (0.15, 4.08)	.77	—
Seminal vesicle†	1.88 (1.05, 3.37)	.03*	—
Small bowel	1.54 (0.86, 2.76)	.15	—
Colon	1.11 (0.32, 3.86)	.87	—
Peritoneum	1.16 (0.56, 2.44)	.69	—
Posterior structure involvement	1.73 (1.09, 2.74)	.02*	—
Most cranial extension
Level L5	4.08 (1.00, 16.62)	.05*	—
Level S1	2.44 (1.26, 4.69)	.008*	—
Level S2	0.98 (0.40, 2.36)	.96	—
Level S3	1.07 (0.51, 2.26)	.85	—
Level S4	0.64 (0.20, 2.03)	.45	—
Level S5	0.61 (0.33, 6.73)	.61	—
Level coccyx	0.86 (0.26, 2.87)	.81	—
Presacral fascia	1.75 (1.10, 2.79)	.02*	1.33 (0.78, 2.26)	.30
Periosteum/cortex	2.69 (1.47, 4.91)	.001*	2.25 (1.14, 4.47)	.02*
Bone marrow	1.44 (0.56, 3.69)	.45	—
Lateral structure involvement	2.00 (1.25, 3.19)	.004*	—
Sciatic nerve	1.04 (0.54, 2.00)	.91	—
Sacral nerve root
S1-S2 nerve root	1.24 (0.47, 3.28)	.67	—
S3-S5 nerve root	1.47 (0.74, 2.93)	.27	—
Lumbosacral plexus	0.82 (0.34, 1.99)	.66	—
Sacrospinous ligament	1.36 (0.83, 2.22)	.22	—
Piriformis muscle	1.88 (0.95, 3.72)	.07	1.48 (0.73, 3.03)	.28
Obturator muscle	1.81 (1.01, 3.23)	.046*	1.45 (0.78, 2.68)	.24
Vascular	2.00 (1.21, 3.33)	.007*	1.73 (1.00, 2.98)	.05*
Common iliac artery	1.98 (0.49, 8.08)	.34	—
Internal iliac artery	1.99 (1.12, 3.56)	.02*	—
External iliac artery	2.65 (0.58, 12.07)	.21	—
Common iliac vein	2.41 (0.72, 8.08)	.15	—
Internal iliac vein	2.21 (1.29, 3.79)	.004*	—
External iliac vein	2.00 (0.63, 6.35)	.24	—

^*Statistically significant.

^†Seminal vesicle involvement was omitted in the multivariable analysis due to collinearity.

Association of Anatomic Structure Involvement with Resection Margin Status

Within the central compartment, the recurrence was located at the anastomosis in 83.7% (164 of 196) of patients with central compartment involvement (Table 2), and univariable analysis revealed that only involvement of the anastomosis was significantly associated with lower risk of R1 resection (OR, 0.52; 95% CI: 0.32, 0.84; P = .008), rendering it the driving factor within the central compartment. In the anterior compartment, both involvement of the prostate and ureter were significantly associated with an increased risk (OR, 2.01; 95% CI: 1.06, 3.80; P = .03 and OR, 2.43; 95% CI: 1.43, 4.13; P = .001, respectively). At multivariable analysis, only ureter involvement was associated with R1 resections (OR, 2.46; 95% CI: 1.44, 4.19; P < .001). Within the posterior and lateral compartments, extension of the tumor cranially from S1 (OR, 2.44; 95% CI: 1.26, 4.69; P = .008) and upward to L5 (OR, 4.08; 95% CI: 1.00, 16.62; P = .05), as well as vascular involvement (OR, 2.00; 95% CI: 1.21, 3.33; P = .007) and obturator muscle involvement (OR, 1.81; 95% CI: 1.01, 3.23; P = .046) were significantly associated with increased likelihood of R1 resection. Multivariable analysis demonstrated that involvement of the periosteum and/or bony cortex (OR, 2.25; 95% CI: 1.14, 4.47; P = .02) and vascular involvement (OR, 1.73; 95% CI: 1.00, 2.98; P = .05) were also associated with higher likelihood of R1 resection, making them driving factors within their respective compartments (Table 3).

Association of Compartment Involvement with Oncologic Outcomes

Involvement of the central compartment was associated with better OS (HR, 0.76; 95% CI: 0.59, 0.99; P = .045) and DFS (HR, 0.69; 95% CI: 0.53, 0.90; P = .006). Involvement of the anterior compartment (HR, 1.41; 95% CI: 1.07, 1.87; P = .01) was associated with worse DFS. Posterior compartment involvement was not associated with OS (HR, 1.26; 95% CI: 0.97, 1.63; P = .09) or DFS (HR, 1.27; 95% CI: 0.98, 1.65; P = .07). Lateral compartment involvement led to worse OS (HR, 1.36; 95% CI: 1.05, 1.77; P = .02) and DFS (HR, 1.49; 95% CI: 1.14, 1.94; P = .003). Even among patients in whom an R0 resection was achieved, central recurrences were associated with improved DFS (HR, 0.69; 95% CI: 0.49, 0.99; P = .045), whereas lateral recurrences were associated with worse DFS (HR, 1.51; 95% CI: 1.07, 2.12; P = .02), as shown in Figures 3 and 4.

Figure 3:

Graph shows disease-free survival of patients with radical resection (R0) of lateral and nonlateral locally recurrent rectal cancer (LRRC).

Figure 4:

Graph shows disease-free survival of patients with radical resection (R0) of central and noncentral locally recurrent rectal cancer (LRRC).

Association of Tumor Morphology with Resection Margin Status

The odds of positive resection margins were influenced by tumor size, tumor type, and border type (Figs 5 and 6). Tumor size was inversely associated with resectability. Larger tumors were associated with increased likelihood of an R1 resection (OR, 1.12; 95% CI: 1.02, 1.24; P = .02) (Table 4). Mucinous tumors were associated with reduced risk of R1 resection compared with other tumor types (OR, 0.39; 95% CI: 0.19, 0.78; P = .008). On the contrary, solid tumors were associated with increased risk (OR, 1.91; 95% CI: 1.09, 3.36; P = .02) (Fig 3). Furthermore, tumors with a spiculated or irregular border were associated with a higher risk of R1 resection (OR, 2.71; 95% CI: 1.41, 5.19; P = .003) (Fig 4).

Figure 5:

T2-weighted MR images of tumor types in locally recurrent rectal cancer. (A) Non–contrast-enhanced T2-weighted transverse MR image of a well-defined tumor border in a 73-year-old-male patient. (B) Non–contrast-enhanced T2-weighted transverse MR image of a spiculated or irregular tumor border in an 88-year-old male patient. Tumors are circled in blue.

Figure 6:

T2-weighted MR images of border types in locally recurrent rectal cancer. (A) Non–contrast-enhanced T2-weighted transverse MR image of central compartment involvement in a 67-year-old male patient. (B) Non–contrast-enhanced T2-weighted transverse MR image of anterior compartment involvement in a 77-year-old male patient. (C) Non–contrast-enhanced T2-weighted sagittal MR image of posterior compartment involvement in a 46-year-old male patient. (D) Non–contrast-enhanced T2-weighted transverse MR image of lateral compartment involvement in a 58-year-old male patient. Tumors are circled in blue.

Table 4:

Association of MRI-derived Tumor Characteristics of Patients with Locally Recurrent Rectal Cancer with Surgical Resection Margins

Tumor Characteristic	Univariable Logistic Regression
	Odds Ratio*	95% CI	P Value
Tumor size (per 1-cm growth)	1.12	1.02, 1.24	.02†
Tumor type
Mucinous	0.39	0.19, 0.78	.008†
Solid	1.91	1.09, 3.36	.02†
Fibrotic	0.97	0.40, 2.34	.95
Tumor border type
Well-defined	0.37	0.19, 0.71	.003†
Spiculated or irregular	2.71	1.41, 5.19	.003†

^*For irradical resection.

^†Statistically significant.

Association of Tumor Morphology with Oncologic Outcomes

OS and DFS were not independently influenced by tumor size, tumor type, and border type, except for fibrotic tumors, which demonstrated a decreased DFS (HR, 1.63; 95% CI: 1.02, 2.62; P = .04) (Table 5). An irregular or spiculated border type (HR, 1.33; 95% CI: 0.96, 1.85; P = .09) and a well-defined border type (HR, 0.75; 95% CI: 0.54, 1.05; P = .09) were not associated with DFS. When corrected for resection margin status, fibrotic tumors (HR, 1.66; 95% CI: 0.91, 3.01; P = .09) and solid tumors (HR, 0.72; 95% CI: 0.50, 1.04; P = .08) did not retain their association with DFS.

Table 5:

Association of MRI-derived Tumor Characteristics of Patients with Locally Recurrent Rectal Cancer with Overall Survival and Disease-Free Survival

Tumor Characteristic	Overall Survival			Disease-Free Survival
	HR	95% CI	P Value	HR	95% CI	P Value
Tumor size (per 1-cm growth)	1.02	0.96, 1.08	.50	0.98	0.93, 1.04	.98
Tumor type
Mucinous	1.07	0.77, 1.50	.68	1.01	0.72, 1.43	.94
Solid	0.87	0.65, 1.18	.38	0.84	0.62, 1.13	.84
Fibrotic	1.28	0.76, 2.16	.36	1.63	1.02, 2.62	.04*
Tumor border type
Well-defined	0.90	0.65, 1.24	.51	0.75	0.54, 1.05	.09
Spiculated or irregular	1.11	0.81, 1.53	.51	1.33	0.96, 1.85	.09
R0 resected recurrences
Tumor size (per 1-cm growth)	1.00	0.92, 1.08	.98	0.94	0.86, 1.02	.14
Tumor type
Mucinous	1.28	0.86, 1.90	.22	1.22	0.81, 1.83	.35
Solid	0.75	0.52, 1.08	.12	0.72	0.50, 1.04	.08
Fibrotic	1.32	0.67, 2.61	.42	1.66	0.91, 3.01	.10
Tumor border type
Well-defined	1.02	0.69, 1.48	.94	0.86	0.58, 1.28	.46
Spiculated or irregular	0.99	0.67, 1.44	.94	1.16	0.78, 1.73	.46
Involved compartment
Central	0.76	0.59, 0.99	.045*	0.69	0.53, 0.90	.006*
Anterior	1.23	0.94, 1.62	.14	1.41	1.07, 1.87	.01*
Posterior	1.26	0.97, 1.63	.09	1.27	0.98, 1.65	.07
Lateral	1.36	1.05, 1.77	.02*	1.49	1.14, 1.94	.003*

Note.—Data were calculated using Cox regression analysis. HR = hazard ratio, R0 = radical resection.

^*Statistically significant.

Discussion

The objective of this study was to evaluate the influence of MRI-detected tumor invasion of surrounding pelvic structures and tumor morphology on resectability of LRRC and postoperative oncologic outcomes in 328 patients. Analyses demonstrated that central compartment involvement facilitated improved resectability and oncologic outcomes, whereas lateral compartment involvement was associated with increased risk of R1 resection and worse oncologic outcome. Improved DFS for central recurrences and worse DFS for lateral recurrences was also found after adjusting for resection margin status. Additionally, tumor size, tumor type, and border type all influenced the resectability of LRRC, and both fibrotic tumors and an irregular or spiculated tumor border type were associated with decreased DFS. After correcting for resection margin status, there was no significant association between solid and fibrotic tumors with DFS.

As previously reported, central recurrences have the most favorable survival-related outcomes and resectability (6–12). The majority of central recurrences in this cohort were located at the anastomosis. The higher odds of an en bloc resection with negative resection margins of central recurrences may be explained by the possibility to achieve wider resection margins due to more soft-tissue surroundings. Furthermore, R0-resected central recurrences demonstrate less local-regional and/or distant recurrence compared with other R0-resected recurrences, which may suggest a less aggressive biologic behavior. Few studies describe the likely etiology of central recurrences, despite its potential influence on the prognosis. It has been suggested that anastomotic recurrences occur due to perioperative tumor spillage, through an anastomotic leakage, remnant tumor cells after resection, or due to an incomplete total mesorectal excision (6,24,25). This implies that central recurrences occur as a consequence of technical shortcomings, rather than unfavorable biologic features. These technical shortcomings can occur in all cases of rectal cancer, specifically also in cases of early-stage, non-neoadjuvantly treated, biologically favorable tumors, possibly explaining the better prognosis.

Anterior compartment involvement overall did not influence resectability. However, involvement of the prostate and ureter were associated with an increased likelihood of R1 resections. Tumor extension cranially from S1 and upward also significantly influenced the odds of positive resection margins. The fact that involvement of these structures has a significant influence on the resectability, independent of involved compartments, supports standardized reporting of anatomic structures within all compartments.

Lateral recurrences are associated with poor oncologic outcomes and a reduced likelihood of radical resection when compared with all other recurrences (6–12). The poor resectability may be a result of the more technically challenging tumor location, surrounded by vital neurovascular structures and the bony pelvis, resulting in challenging procedures and higher R1-resection rates. In a study by Westberg et al (12), lateral recurrences were associated with a worse OS compared with other recurrences, even when corrected for radicality. In our study, there was no increased likelihood of death after R0 resection of lateral recurrences compared with nonlateral recurrences, but the recurrence rate of lateral recurrences was significantly worse in comparison to other recurrences, regardless of resection margin status. This suggests a more aggressive biologic behavior. Many studies suggest that lateral recurrences occur as a consequence of lateral pelvic lymph node involvement, which are either not recognized or not resected (6,26,27). However, MRI features of primary rectal cancer characteristics, such as extramural venous invasion and tumor deposits, have also been described to increase the risk of local recurrence at the resection margins and the lateral compartment (18–20,28). These characteristics are thought to reflect poor biologic behavior, because tumors with these characteristics have already demonstrated a tendency to spread to and/or invade surrounding tissue and structures and increase the risk of distant metastases. All these factors combined support the hypothesis that lateral recurrences reflect poor tumor biology with an unfavorable prognosis. Primary rectal tumors with any of these characteristics are routinely subjected to neoadjuvant treatment, which could result in local recurrences arising from therapy-resistant cells with a poorer prognosis.

As reported previously, involvement of pelvic structures as well as tumor size, tumor type, and border type play an important role in determining the resectability of LRRC and patient survival (12). Larger tumors are more likely to invade into surrounding structures, thereby increasing the likelihood of an R1 resection. However, Cox proportional hazard models, calculated per 1-cm tumor size increase, demonstrated no decreased survival-related outcomes for larger tumors. Nevertheless, due to the limited room within the lesser pelvis, expansion of LRRC quickly also results in crossing the boundaries of the pelvic compartments, resulting in the involvement of multiple compartments. However, identifying the primary involved compartment at MRI seems to provide an indication of the etiology and the prognosis. Solid tumors were, contrary to our expectation, associated with more R1 resections compared with the other tumor types. This did, however, not translate into survival differences. In fact, fibrotic tumors demonstrated a significantly worse DFS compared with solid tumors, despite demonstrating a significantly better resectability than solid tumors. Additionally, DFS was also negatively influenced in tumors with an irregular or spiculated border. Therefore, both characteristics could be considered reflections of unfavorable biologic behavior. In all R0-resected recurrences, fibrotic tumors and solid tumors did not retain a significant association with DFS. Although no conclusions should be drawn from these outcomes, this might support the notion of a possible difference in biologic behavior.

Our data on the involvement of different compartments and tumor morphology offer additional insight into the prognosis of patients with LRRC and the variation in biologic behavior of LRRC. The data support the use of multidisciplinary discussions to determine which patients are less suitable for surgical treatment after considering tumor characteristics. Selection of patients with favorable biologic behavior could improve the prognosis and reduce the number of R1 resections in patients with LRRC. On the contrary, refraining from extensive radical resections in patients with a poor prognosis because of poor tumor biology, regardless of whether complete surgical resection seems feasible or not, results in the prevention of unnecessary surgery that does not influence oncologic outcome. Overall, this study demonstrates that adopting a standardized reporting checklist could lead to a more systematic approach in collecting data, which could potentially increase our understanding of LRRC tumor characteristics and the prognosis of LRRC and lead to better patient selection.

Our study had several important limitations. The main limitation of this study is the risk of selection bias, as all participating patients were treated surgically, and data on patients who did not undergo surgery or who were treated with palliative intent were absent from the analysis. Furthermore, the study period from 2003 to 2017 saw significant advancements in MRI technology. Throughout this time, both centers consistently used a 1.5-T MRI as the standard, ensuring a stable and reliable imaging baseline. In the later years, there was a shift toward the increased use of 3-T MRI, especially following the release of the European Society of Gastrointestinal and Abdominal Radiology guidelines, which introduced enhanced quality standards for clinical imaging. These improvements, particularly the transition to 3-T MRI, could have contributed to improved resolution and diagnostic accuracy. However, it is important to note that all MRI scans performed during the study were evaluated by experienced radiologists, who deemed them to be of sufficient quality for clinical assessment, in accordance with their professional discretion. This approach ensures that technological variations over the study period did not significantly affect the study’s outcomes. However, the differences in MRI quality may have resulted in incomplete reporting. Furthermore, the duration of the study period may have had an impact on the outcome of patients. Although the neoadjuvant and surgical treatment of LRRC has remained similar, incremental advancements in different aspects of the treatment probably have led to improved patient outcomes since the start of the study period, which must be taken into account.

In conclusion, our study demonstrates that tumor involvement of certain pelvic compartments, as well as specific MRI-derived characteristics, such as a fibrotic tumor type or spiculated or irregular tumor borders, are associated with a worse prognosis in patients with LRRC. Moreover, lateral compartment involvement remains associated with worse prognosis regardless of resection margin status. Therefore, these tumor characteristics may be considered as negative prognosticators and elicit a cautious approach in curative treatment. Standardizing radiology reports for pelvic MRI for LRRC may facilitate more optimal patient selection and provide further insight into prognosticators of LRRC. Prospective studies collecting standardized MRI data on LRRC are needed to confirm our findings and identify further relevant prognosticators.