Type of recurrence is associated with disease-free survival after salvage surgery for locally recurrent rectal cancer

Abstract

Purpose

To compare the characteristics and outcomes of rectal cancer patients with local recurrence at a perianastomotic site (PA), a surgical field (SF) site, or in lateral lymph nodes (LLN).

Methods

A total of 114 consecutive patients who underwent surgery for recurrent, non-metastatic rectal cancer at a single comprehensive cancer center between 1997 and 2012 were grouped on the basis of radiographic assessment of type of recurrence: PA, 76 (67%) patients; SF, 25 (22%) patients; LLN, 13 (11%) patients. Demographic, clinical, and pathological features were compared between the three groups, as were disease-free survival (DFS) and overall survival (OS).

Results

Recurrence type was associated with positive circumferential margin in the primary resection (PA, 4 [6%]; SF, 4 [19%]; LLN, 3 [25%]; P = 0.027), prior neoadjuvant therapy for the primary tumor (PA, 57 [75%]; SF, 18 [72%]; LLN, 4 [31%]; P = 0.007) and location of the primary tumor in the upper rectum (PA, 33 [45%]; SF, 5 [23%]; LLN, 1 [8%]; P < 0.001). Patients with PA had longer median DFS (PA, 5.1 years; SF, 1.5 years; LLN, 1.2 years; P = 0.036). There was a non-significant trend toward longer OS and higher rates of R0 resection for PA.

Conclusion

Type of recurrence after salvage surgery for locally recurrent rectal cancer is associated with longer DFS in patients with PA recurrence.

Introduction

Recurrent rectal cancers are associated with poor prognosis, and many patients are inoperable at the time of diagnosis due to extensive distant metastasis or advanced local spread. For patients who are candidates for surgery, potential benefits must be weighed against the morbidity and mortality of repeat surgery [1, 2]. With the advent of total mesorectal excision [3] and multimodal treatment combining radiation and chemotherapy [4–6], the incidence of local recurrence has decreased from 31% to 10% [3, 7]. The patterns of local recurrence have also changed, with increased incidence of non-central recurrences compared to the pre-TME era [6, 8].

Previous studies have identified multiple risk factors for local recurrence of rectal cancer, including stage IV, T4, N2, and positive circumferential resection margin in T3 disease [9–11]. However, whether these risk factors also predict site of recurrence has not been established. A better understanding of this relationship may help decrease local recurrence rates by modification of surgical technique or through the introduction of new practice parameters, such as radiotherapy field designs, extended lymphadenectomies, and consideration of adjuvant therapy for early stage disease.

Multiple frameworks have been developed for categorizing recurrent rectal cancer on the basis of tumor location. One of the first was proposed in 1996 by Suzuki et al. [12]. It defined four locations—anterior, right, posterior, and left—with up to three points of fixation for each location. Patients with more extensive fixation had more complications following resection, but the relationship between site of recurrence and survival was not clearly defined. Wanebo et al. [13] developed a classification schema with five categories ranging from intramural local recurrence to infiltration of the bone or ligaments. The latter category was associated with the need for aggressive resection, such as sacrectomy in conjunction with abdominoperineal resection. Yamada et al. [14] categorized recurrence in the pelvis as localized, sacral, or lateral and found an association between these categories and survival, with 5-year overall survival ranging from 38% for localized to 10% for sacral and 0% for lateral. Similarly, Belli at el. [15] categorized recurrence according to location and degree of invasion of bony structures of the pelvis and found that lateral pelvic wall involvement was associated with positive resection margins and poor survival.

Prior classification systems of recurrent rectal cancer were primarily based on location of recurrence and degree of invasion of nearby pelvic structures, for assessment of the potential for resection with negative margins, since positive margins are associated with poor prognosis [2, 16, 17]. Although previous classifications provide important information on the management of patients with recurrence, they did not categorize recurrences based on pathophysiological mechanism. Our aim was to develop a simple yet practical classification system for rectal cancer recurrence based on an assessment of preoperative imaging that reflects three different pathophysiological types of recurrence: perianastomotic recurrence (PA), surgical field recurrence (SF) and lateral lymph node recurrence (LLN). Our hypothesis was that these three types of recurrences would be associated with different prognoses.

Methods

With a waiver of informed consent and approval of the institutional review board, a prospectively maintained database was queried to identify patients who underwent surgery for local recurrence of rectal cancer at Memorial Sloan Kettering Cancer Center between January 1, 1997, and December 31, 2012. Local recurrence was defined as a single tumor lesion in the pelvis or perineum after an R0/R1 curative resection of the primary tumor, as confirmed by pathological analysis of a preoperative biopsy sample or the surgical specimen.

The data retrieved included the date of the operation for the primary rectal cancer and clinicopathologic findings. Only mesorectal excision for primary tumors were included. Local or endoscopic resection were excluded. Positive margin was defined as presence of tumor cells within 1 mm of the margin. Location of the primary tumor was categorized on the basis of the initial endoscopic examination as follows: lower rectum, <6 cm from the anal verge; middle rectum, ≥ 6 cm to ≤ 12 cm from the anal verge; upper rectum, > 12 cm to ≤ 15 cm from the anal verge. Patients with distant or multifocal recurrences were excluded. OS and DFS were calculated from the date of surgery. Events for DFS included local recurrence, distant recurrence and death.

Each patient’s first available MRI or CT scan for which the attending radiologist’s official report indicated local recurrence was re-reviewed by an expert radiologist (M.J.G.), blinded to the clinical data. Local recurrence was defined radiologically as any infiltrative, expansive, or asymmetrically located pelvic mass with or without contrast enhancement that could not be explained by postoperative changes.

Recurrence types

Type of recurrence was classified as follows (Fig. 1). PA was defined as a predominantly (>50%) intramural lesion with an involved staple line or a predominantly (>50%) extramural lesion in the mesorectum, at or below the level of the anastomosis. LLN was defined as lateral recurrence in the obturator lymph node compartment or along the common or internal iliac vessels. SF was defined as recurrence at any other location including (i) posterior (predominantly [>50%] midline posterior lesion, outside the neo-mesorectum [e.g., presacral]), (ii) anterior (predominantly [>50%] midline anterior lesion, outside the neo-mesorectum), (iii) lateral (predominantly [>50%] lateral lesion, outside the mesorectum, other than in a lateral lymph node), or (iv) perineal (predominantly perineal lesion, including lesions in the anal sphincter complex and the surrounding perianal and ischiorectal space).

Figure 1.

Sample MRI/CT images of perianastomotic (upper panel), lateral lymph node (middle panel), and surgical field (lower panel) recurrences.

Statistical analysis

Statistical analysis of clinicopathologic variables was performed using SAS software version 9.4. Categorical variables were compared using Fisher’s exact test, and continuous variables were compared using the two-tailed t-test with the Bonferroni correction for multiple comparisons. Kaplan-Meier estimates and log-rank tests were used to analyze differences in disease-free survival and overall survival. P values were two-sided, and P < 0.05 was considered statistically significant.

Results

One hundred forty-seven consecutive patients underwent surgery for local recurrence of rectal cancer at Memorial Sloan Kettering during the study period. All 147 patients had initially undergone an intended curative resection (anterior resection, low anterior resection, abdominoperineal resection, or Hartmann’s procedure after removal of the primary tumor) for primary rectal adenocarcinoma within 15 cm of the anal verge. Thirty-three of the 147 patients were excluded from the study because of stage IV disease, multiple recurrence sites, or local recurrence from another primary tumor (Fig. 2). Of the 114 patients included in the study, 61 (53.5%) were men. Median age at the time of primary surgery was 57 years. The median interval from primary surgery to local recurrence was 20 months. The median follow-up period after salvage surgery for recurrence was 4.4 years (interquartile range, 2.1–7.2 years).

Figure 2.

Inclusion and grouping of patients.

Clinicopathologic and treatment characteristics of patients with PA, SF or LLN recurrence are listed in Table 1. PA was the most common recurrence type (76 of 114 patients; 67%), followed by SF (25 of 114 patients; 22%) and then LLN (13 of 114 patients; 11%).

Table 1.

Clinicopathologic and treatment characteristics (n = 114) in relation to the site of recurrence

Characteristic	No. of patients (%)			P a
	Perianastomotic recurrence (n = 76)	Surgical field recurrence (n = 25)	Lateral lymph node recurrence (n = 13)
Age, mean ± SD	57.3 ± 11.4	56.2 ± 12.8	48.9 ± 8.9	0.057
Sex				0.391
Female	32 (42)	14 (56)	7 (54)
Male	44 (58)	11 (44)	6 (46)
Location of primary tumor (n = 107)*				<0.001
Lower rectum	12 (16)	12 (55)	7 (58)
Middle rectum	28 (38)	5 (23)	4 (33)
Upper rectum	33 (45)	5 (23)	1 (8)
Neoadjuvant therapy for primary tumor*				0.008
None	19 (25)	7 (28)	9 (69)
CT-CRT	57 (75)	18 (72)	4 (31)
pT category (n = 111)*				0.580
0	2 (3)	0	0
1	9 (12)	5 (21)	1 (8)
2	20 (27)	5 (21)	4 (31)
3	41 (55)	11 (46)	8 (62)
4	2 (3)	3 (13)	0
pN category*				0.130
0	51 (67)	19 (76)	5 (38)
1	18 (24)	3 (12)	5 (38)
2	7 (9)	3 (12)	3 (23)
Circumferential resection margin (n = 102)*				0.027
Negative	65 (94)	17 (81)	9 (75)
Positive	4 (6)	4 (19)	3 (25)
Distal resection margin (n = 96)*				0.909
≤2 cm	32 (52)	11 (50)	7 (58)
>2 cm	30 (48)	11 (50)	5 (42)
Differentiation extent (n = 109)*				0.152
Well differentiated	2 (3)	4 (16)	0 (0)
Moderately differentiated	60 (83)	17 (68)	11 (92)
Poorly differentiated	10 (14)	4 (16)	1 (8)
Lymphovascular invasion (n = 85)*				0.119
No	41 (69)	14 (88)	5 (50)
Yes	18 (31)	2 (12)	5 (50)
Adjuvant therapy for primary tumor*				0.111
None	28 (37)	8 (32)	1 (8)
CT-CRT	48 (63)	17 (68)	12 (92)
Hospital of primary treatment *				0.013
Other than MSKc	60 (79)	19 (76)	5 (38)
MSK	16 (21)	6 (24)	8 (62)
Neoadjuvant therapy for recurrence				0.324
None	21 (28)	9 (36)	6 (46)
CT-CRTb	55 (72)	16 (64)	7 (54)
Intraoperative radiotherapy for recurrence				0.248
No	18 (24)	2 (8)	3 (23)
Yes	58 (76)	23 (92)	10 (77)
Surgical procedure for recurrence
Abdominoperineal resection	6 (8)	9 (36)	4 (31)	0.005
Sphincter-preserving resection	66 (87)	15 (60)	8 (62)
Other	4 (5)	1 (4)	1 (8)
Resection margins of recurrence
R0	55 (72)	13 (52)	7 (54)	0.111
R1/R2	21 (28)	12 (48)	6 (46)
Vital status (n = 117)				0.0206
No evidence of disease	31 (41)	4 (16)	1 (8)
Alive with disease	4 (5)	2 (8)	2 (15)
Died of disease	40 (53)	19 (76)	10 (77)

^aFisher’s exact test (categorical variables) or a two-tailed t-test with the Bonferroni correction for multiple comparisons (continuous variables). P values < 0.05 are boldfaced.

^bCT-CRT, systemic chemotherapy and chemoradiotherapy.

^cMSK, Memorial Sloan Kettering Cancer Center.

^*Data for primary tumor and procedure.

Primary tumor location information was available for 107 of 114 patients (94%). Primary tumor in the upper rectum was associated with PA (PA, 33 of 73 patients with available information [45%]; SF, 5 of 22 patients [23%]; LLN, 1 of 12 patients [8%]; P < 0.001). Data on circumferential resection margin of the primary tumor was available for 102 of 114 patients (89%). SF and LLN were associated with patients with positive circumferential resection margin (PA, 4 of 69 patients with available information [4%]; SF, 4 of 21 patients [19%]; LLN, 3 of 12 patients [25%]; P = 0.027). LLN was associated with patients who did not receive neoadjuvant treatment for their primary tumor (PA, 19 of 76 patients [25%]; SF, 7 of 25 patients [28%]; LLN, 9 of 13 patients [69%]; P = 0.007). PA and SF were associated with primary surgery performed outside our medical center (PA, 60 of 73 patients [79%]; SF, 19 of 25 patients [76%]; LLN, 5 of 13 patients; P = 0.013). Sex, T stage, N stage, tumor differentiation, lymphovascular invasion, and distal resection margins did not differ significantly between the three groups of patients with different sites of recurrence. The three groups also did not differ significantly in having received adjuvant therapy following primary surgery.

Patients who underwent salvage surgery for PA were more likely to undergo a sphincter-preserving operation (PA, 66 of 76 patients [87%]; SF, 15 of 25 patients [60%]; LLN 8 of 13 patients [62%]; P = 0.005). Patients with PA also had a non-significant trend toward a higher rate of R0 re-resection (PA, 55 of 76 patients [72%]; SF, 13 of 25 patients [52%]; LLN, 7 of 13 patients [54%]; P = 0.111). With median follow-up of 4.4 years after salvage surgery, 31 of 76 patients (41%) who underwent surgery for PA were alive without disease, compared with 4 of 25 patients (16%) who underwent surgery for SF and 1 of 13 patients (8%) who underwent surgery for LLN recurrence (P = 0.02).

Patients with PA had longer median disease-free survival (5.14 years) than patients with SF (1.52 years) or LLN (1.16 years) (P = 0.03; Fig. 3). Overall survival followed a similar pattern (6.13, 4.43, and 4.40 years, respectively), but the difference was not statistically significant. Patients with an R0 salvage resection had longer median overall survival than patients whose salvage resection was R1 or R2 (6.13, 3.6, and 0.3 years, respectively; P = 0.001; Fig. 4).

Figure 3.

Disease-free survival (A) and overall survival (B) from the date of salvage surgery for patients with perianastomotic, surgical field, or lateral lymph node recurrence, compared using the log-rank test.

Figure 4.

Overall survival from the date of salvage surgery for patients whose salvage surgery was R0, R1, or R2. P = 0.01 (log-rank test).

Discussion

In this study, we utilized a new classification system of different pathophysiological types of recurrence after primary rectal cancer surgery and investigated this relationship with DFS and OS. We defined three types of recurrence: PA, SF and LLN. In contrast to previous reports, the majority of recurrences in our cohort were PA and had primarily an intramural component[18]. Patients with PA had longer median DFS than patients with SF or LLN recurrence. Worse DFS for LLM is consistent with previous evidence showing shorter DFS for tumors infiltrating the lateral pelvic compartment [15]. Interestingly, R0 resection of tumors infiltrating the lateral pelvic compartment may not be associated with a survival benefit [15].

We described associations between primary tumor clinicopathologic and treatment characteristics and recurrence type. PA was associated with tumors in the upper rectum, SF and LLN were associated with positive circumferential resection margin, and LLN was associated with omission of neoadjuvant treatment. It is likely that LLN is due to lateral pelvic lymph node involvement at baseline which occurs primarily in low tumors and could be treated with neoadjuvant chemoradiation [19] in selected cases. Our data suggests that omission of neoadjuvant treatment may be associated with later development of LLN.

Consistent with previous studies [2, 17, 18], salvage operations with R0 resection margins were associated with longer OS in comparison to R1/R2 margins. We also detected a trend toward a higher rate of R0 resection in patients who underwent salvage surgery for PA. Previous reports [15, 20], demonstrated lower rates of R0 resection for recurrences involving the lateral pelvic sidewall.

In some patients whose primary tumor was in the upper rectum, PA may be associated with the remaining mesorectum, as suggested by the findings reported by Syk et al. [6]. In a retrospective study involving 33 patients with local recurrence, they found that all primary tumors from the upper rectum recurred at the anastomosis, with 86% containing evidence of residual mesorectal fat. Their data argued that partial mesorectal excision (PME) may lead to perianastomotic recurrence due to insufficient mesorectum excision. Current guidelines suggest PME as acceptable for upper and middle rectal cancer as long as there is a minimum of 5 cm distal resection of mesorectum [16]. In our cohort, in part because many of the initial operations were not done at Memorial Sloan Kettering, we could not obtain adequate data regarding the length of resected mesorectum to further evaluate this claim.

Syk et al [6] also reported that most presacral recurrences were found in patients whose primary tumor was in the middle rectum, probably due to the difficulty of excision at that level, and only tumors in the lower rectum were associated with recurrence in the pelvic floor. In our study, among patients who had recurrence in the SF, 55% were tumors in the lower rectum (P < 0.001). This association may be explained by the challenge of dissecting the distal rectum and the limited exposure where the surgical technique of forcibly retracting a thin mesorectum may lead to exfoliation of tumor cells, positive circumferential resection margin, or rectal perforation [10,17].

In this study, more than 75% of patients with PA and LLN recurrence and 92% of patients diagnosed with SF recurrence received intraoperative radiotherapy (IORT). There is some controversy regarding the use of IORT for local recurrence after rectal cancer resection. It is clear that to obtain an R0 resection is more important than giving IORT boost for long-term outcomes [21], nevertheless, there are studies showing that IORT could improve long-term oncological results even in re-irradiated patients [22, 23], and that is why this modality is frequently utilized at our institution.

We found differences between patients treated initially at Memorial Sloan Kettering versus other centers, with more PA and SF recurrences in patients referred from outside centers. This raises the question of the potential benefit of centralization of complex rectal cancer procedures [24, 25]. However, we do not know the denominator of the rectal cancer cases in the outside centers, and the number of patients in our study is too small to make definitive conclusions.

Our study had limitations that are inherent in retrospective studies, including selection bias. All salvage surgeries for recurrence were performed at a single specialized cancer center, which limits generalizability. The analysis of circumferential and distal resection margins and of lymphovascular invasion was hampered by non-standardized documentation. The mesorectum was also difficult to identify on some CT scans, and tumor location could not always be determined precisely. Additionally, we were unable to reliably identify the pre-treatment clinical disease stage for some patients who received neoadjuvant therapy outside Memorial Sloan Kettering.

In summary, using a practical schema for categorizing post-surgery rectal cancer recurrence as PA, SF, or LLN, our study found that following salvage surgery, patients with PA have longer DFS. Our new classification system should be further validated in future prospective studies to better elucidate the pathophysiological determinants of local recurrence of rectal cancer.