Endoscopic Predictors of Residual Tumor After Total Neoadjuvant Therapy: A Post Hoc Analysis From the Organ Preservation in Rectal Adenocarcinoma Trial

Abstract

BACKGROUND:

Restaging endoscopy plays a critical role in selecting locally advanced rectal cancer patients who respond to neoadjuvant therapy for nonoperative management.

OBJECTIVE:

This study evaluated the restaging endoscopic features that best predict the presence of residual tumor in the bowel wall.

DESIGN:

This was a post hoc analysis of a prospective randomized trial.

SETTINGS:

The Organ Preservation in Rectal Adenocarcinoma Trial randomized patients across 18 institutions with stage II/III rectal adenocarcinoma to receive either induction or consolidation total neoadjuvant therapy. Surgeons completed a restaging tumor assessment form, which stratified patients across three tiers of clinical response.

PATIENTS:

Patients enrolled in the Organ Preservation in Rectal Adenocarcinoma Trial with a completed tumor assessment form were included.

MAIN OUTCOME MEASURES:

The main outcome was residual tumor, which was defined as either an incomplete clinical response or local tumor regrowth within two years of restaging. Independent predictors of residual tumor were identified using backwards-selected multivariable logistic regression analysis. Subgroup analyses for complete and near complete clinical responders were performed.

RESULTS:

Surgeons completed restaging forms for 263 patients at a median of 7.7 weeks after neoadjuvant therapy; 128 (48.7%) had residual tumor. On multivariable regression analysis, several characteristics of a near complete response, including ulcer (OR 6.66; 95%CI 2.54–19.9), irregular mucosa (OR 3.66; 95% CI 1.61–8.68) and nodularity (OR 2.96; 95%CI 1.36–6.58) remained independent predictors of residual tumor. A flat scar was associated with lower odds of harboring residual disease (OR 0.32; 95% CI 0.11–0.93) for patients categorized as complete clinical responders.

LIMITATIONS:

Limitations of this study include analysis of endoscopic features at a single time point and ambiguities in tumor assessment form response criteria.

CONCLUSIONS:

Patients with ulcer, nodularity or irregular mucosa on restaging endoscopy have higher odds of residual tumor. Recognizing these features’ negative prognostic implications will help surgeons better select candidates for nonoperative management and suggests that patients with high-risk characteristics would benefit from close interval surveillance. See Video Abstract at http://links.lww.com/DCR/Bxxx.

INTRODUCTION

The watch-and-wait (WW) approach for locally advanced rectal cancer (LARC) is an accepted alternative to total mesorectal excision (TME) for patients who achieve a complete clinical response (cCR) after neoadjuvant therapy. When applied in the appropriate setting, organ preservation produces equivalent oncologic outcomes and allows patients to avoid the morbidity and long-term functional deficits associated with surgical resection.^1–5 Determining which patients have a cCR relies on a comprehensive restaging assessment several weeks after neoadjuvant therapy consisting of magnetic resonance imaging (MRI), digital rectal exam (DRE) and flexible sigmoidoscopy.^6–9 While combining these diagnostic techniques improves accuracy, a cCR does not always correspond to a pathologic complete response.^10,11 Additionally, endoscopic and radiologic images with mild abnormalities can be difficult to interpret, as these variations can indicate either ongoing tumor regression or residual disease.^12,13

Endoscopy is the most accurate tool available to evaluate a tumor’s luminal response to treatment both at restaging and during WW surveillance.^9,14 However, the exact endoscopic criteria for selecting appropriate WW candidates remain elusive and the implications of specific features poorly described. Overly restrictive definitions of a luminal cCR exclude many patients without residual disease from pursuing nonoperative management (NOM).^15–18 In contrast, allowing patients with minor mucosal abnormalities, or a near complete clinical response (nCR), to proceed with WW may delay definitive treatment for persistent disease or confer a greater risk of local tumor regrowth.^12,19

The Organ Preservation in Patients with Rectal Adenocarcinoma (OPRA) Trial developed the first prospectively collected, standardized method for evaluating endoscopic characteristics of LARC patients treated with total neoadjuvant therapy (TNT).⁴ At the first restaging assessment following TNT, surgeons selected from a discrete set of endoscopic features that correlated to a three-tiered grading system of a cCR, nCR or incomplete clinical response (iCR). Our group has previously reported on the oncologic outcomes of the entire cohort, and demonstrated that the TAF grading system correlates to rates of organ preservation as well as disease-free survival.^4,20 This posthoc analysis of the OPRA trial evaluates the association between individual restaging endoscopic characteristics and residual tumor.

MATERIALS AND METHODS

Patients

The OPRA trial randomized patients 18 and older with biopsy-proven clinical stage II or III rectal adenocarcinoma to receive induction (INCT-CRT) or consolidation (CRT-CNCT) total neoadjuvant therapy. Within 8 ± 4 weeks of completing TNT, patients underwent a restaging assessment consisting of MRI, DRE and flexible sigmoidoscopy. All flexible sigmoidoscopies and DREs were performed at the same appointment by a board-certified colorectal surgeon. Patients determined to have an iCR underwent TME, while those with a cCR proceeded to a standardized WW protocol. Patients with a nCR were eligible either for TME or WW based on the recommendations of their primary treatment team. Details of the inclusion criteria, TNT regimens, WW surveillance protocol and outcomes of the OPRA trial have previously been reported.^4,21 The study protocol was approved by institutional review boards at all participating institutions.

Endoscopic Tumor Assessment Forms

The patient’s primary surgeon completed a standardized endoscopic tumor assessment form (TAF) at restaging. Developed through the expert consensus of all providers participating in the OPRA trial, this form assigned endoscopic and DRE characteristics across three tiers of clinical response (Supplementary Figure 1 at https://links.lww.com/DCR/CXX). Patients with visible or palpable tumor on DRE were classified as an iCR. Those with ulcers, nodularity, mucosal irregularity, mild erythema of the scar or induration on DRE were categorized as a nCR. A cCR was defined as 1) a patient without any of the concerning characteristics listed above 2) endoscopic features such as a flat white scar or telangiectasia and 3) a normal DRE. Patients with characteristics across multiple tiers were assigned to the lowest response group. Figure 1 demonstrates representative endoscopic images from each of the clinical response categories.

Figure 1.

Representative endoscopic images of a complete clinical (cCR), near complete (nCR) and incomplete (iCR) response

a. A flat white scar with telangiectasia demonstrating a cCR

b. A small, well demarcated ulcer without other polypoid growth illustrating a nCR

c. Gross, residual tumor indicative of an iCR

Cohort Definition

Patients were divided into two groups based on response of the primary rectal tumor to TNT. The residual tumor (RT) group included patients with a pathologically proven iCR or local tumor regrowth within 2 years of restaging. The no residual tumor (NRT) group consisted of patients with a pathologic complete response or a sustained cCR for 2 years after restaging. For the purposes of this study, patients with distant metastases who had no remaining disease at the primary rectal tumor site were classified as NRT. Exclusion criteria included 1) patients who did not complete a formal restaging assessment 2) patients with an iCR who refused surgery 3) patients with suspected local tumor regrowth but no pathologic confirmation of residual disease and 4) patients on WW surveillance with less than 2 years of follow up data available.

Statistical Analysis

Demographic, clinical and endoscopic features were denoted as frequencies and proportions for categorical variables. Continuous variables were reported as medians with interquartile range (IQR). Differences across patient characteristics were analyzed using the Wilcoxon rank sum test for continuous variables and Fisher’s exact test for categorical variables. The cut-off value for statistical significance was set as p < 0.05.

The primary outcome of interest was RT. By selecting covariates from a univariate model, a backwards-selected multivariable logistic regression analysis accounting for nodal disease, TNT treatment arm and individual endoscopic features was built to identify independent predictors of RT. This model did not account for DRE findings due to the substantial number of patients without a DRE recorded and an inability to distinguish patients with an unreachable primary tumor from those with a normal exam.

Subset analyses for patients categorized as a cCR or a nCR by TAF criteria were performed to determine whether any endoscopic characteristics predicted RT for these groups. Outcomes were measured by univariate analysis for the cCR group and by a backwards-selected multivariable logistic regression analysis for the nCR cohort. All statistical analyses were performed using R (v. 4.2.2, R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Patient Cohort

Between April 2014 and March 2020, 304 patients randomized in the OPRA trial underwent a restaging assessment. Of these, surgeons completed 294 TAFs. An additional 31 patients were excluded from analysis for not meeting the minimum 2 year follow up requirement, refusal to undergo surgery and lack of pathologically proven local tumor regrowth (Fig. 2). A total of 263 patients were included, with 135 (51.3%) in the NRT group and 128 (48.7%) in the RT group. Among the patients with RT, 66 developed local tumor regrowth at a median of 24 weeks (IQR 31) from restaging assessment. Sixty-five (98.5%) local regrowths occurred intraluminally, and all were salvageable with surgical resection. When classified by TAF clinical response grade, 112 patients met criteria for a cCR, 99 for a nCR and 52 for an iCR.

Figure 2.

Study flow diagram. TNT= total neoadjuvant therapy, TME= total mesorectal excision, TAF= tumor assessment form, cCR= complete clinical response, pCR= pathologic complete response, iCR= incomplete clinical response, LTR = local tumor regrowth

Patients were well matched across baseline characteristics and did not differ by compliance to TNT (Table 1). Patients in the RT group were more likely to have nodal disease at diagnosis (79.7% vs. 63.7%; p = 0.004) and to have received induction TNT (INCT-CRT 56.2% vs. CRT-CNCT 40.7%; p = 0.014). The median interval from end of TNT to restaging assessment was 7.7 weeks and did not differ across groups. None of the patients included in this study had a stoma at the time of restaging assessment.

Table 1:

Baseline and treatment characteristics

Variables	No Residual Tumor (NRT) N=135	Residual Tumor (RT) N=128	p-value
Age (IQR)	59 (18)	57 (17)	0.32
Female gender	50 (37)	38 (29.7)	0.24
Race/Ethnicity			>0.999
White	109 (80.7)	104 (81.2)
Non-White	22 (16.3)	21 (16.4)
Unknown	4 (3)	3(2.4)
cT classification			0.674
cT1 or 2	17 (12.6)	12 (9.4)
cT3	102 (75.6)	99 (77.3)
cT4	16 (11.8)	17 (13.3)
cN positive	86 (63.7)	102 (79.7)	0.004
Tumor distance from AV (IQR), cm	4 (3.8)	4.5 (2.9)	0.775
Treatment arm			0.014
INCT-CRT	55 (40.7)	72 (56.2)
CRT-CNCT	80 (59.3)	56 (43.8)
Received ≥75% of intended neoadjuvant chemotherapy dose	115 (85.2)	107 (83.6)	0.737
Median radiation dose, cGy (IQR)	5400 (580)	5400 (525)	0.158
Received concurrent chemotherapy with radiation	134 (99.3)	128 (100)	>0.999
Median interval TNT to restaging (IQR), weeks	7.8 (3.3)	7.7 (3.8)	0.81

Data are reported as n(%) unless otherwise specified. P-values calculated using Fisher’s exact test for categorical variables and Wilcoxon rank sum for continuous variables. Bolded numbers are statistically significant. Abbreviations: cT classification= clinical tumor classification, cN positive= clinical node status positive, AV= anal verge, INCT-CRT = induction chemotherapy followed by chemoradiation, CRT-CNCT= chemoradiation followed by consolidation chemotherapy

Endoscopic Predictors of Residual Tumor

Patients with RT were more likely to have ulcer (20.3% vs. 4.4%; p < 0.001), nodularity (25.8% vs. 11.1%; p = 0.002), irregular mucosa (26.6% vs. 8.1%; p < 0.001) and visible tumor (32.8% vs. 4.4%; p < 0.001) on restaging endoscopy (Table 2). Flat scar (63.7% vs. 19.5%; p < 0.001) and telangiectasia (50.3% vs. 18%; p < 0.001) were associated with the NRT group. Mild erythema of the scar was not associated with either outcome. On backwards-selected multivariable regression analysis, visible tumor (OR 22.4; 95% CI 9.18–64.1), ulcer (OR 6.66; 95% CI 2.54–19.9), irregular mucosa (OR 3.66; 95% CI 1.61–8.68), nodularity (OR 2.96; 95% CI 1.36–6.58), nodal disease at diagnosis (OR 2.07; 95% CI 1.06–4.17) and INCT-CRT (OR 1.88; 95% CI 1.04–3.44) remained independent predictors of RT (Table 3).

Table 2:

Endoscopic TAF characteristics by tumor response

Variables	No Residual Tumor (NRT) N=135	Residual Tumor (RT) N=128	p-value
Flat scar	86 (63.7)	25 (19.5)	<0.001
Telangiectasia	68 (50.4)	23 (18)	<0.001
Ulcer	6 (4.4)	26 (20.3)	<0.001
Nodularity	15 (11.1)	33 (25.8)	0.002
Mucosal irregularity	11 (8.1)	34 (26.6)	<0.001
Mild erythema of the scar	9 (6.7)	17 (13.3)	0.097
Visible tumor	6 (4.4)	42 (32.8)	<0.001

Data are reported as n (%). P-values calculated using Fisher’s exact test. Bolded numbers are statistically significant.

Table 3:

Multivariable logistic regression: Predictors of residual tumor

Variables	OR	95% CI	p-value
Induction TNT	1.88	1.04–3.44	0.038
cN positive	2.07	1.06–4.17	0.033
Ulcer	6.66	2.54–19.9	<0.001
Nodularity	2.96	1.36–6.58	0.006
Irregular mucosa	3.66	1.61–8.68	0.002
Visible tumor	22.4	9.18–64.1	<0.001

The multivariable logistic regression model was built using backwards selection. Bolded numbers are statistically significant. Abbreviations: OR = odds ratio, 95% CI= 95% confidence interval, TNT= total neoadjuvant therapy, cN positive = clinical node status positive

Endoscopic Features Among Complete Clinical Responders

Of the 112 patients with a cCR, 91 (81.2%) had a flat scar and 74 (66.1%) had telangiectasia on restaging endoscopy. Twenty-three (20.5%) patients developed local tumor regrowth and therefore met criteria for the RT group. Compared to the NRT group, patients with RT did not differ by demographics, treatment arm or tumor characteristics. Univariate logistic regression analysis demonstrated that cCRs with a flat scar had lower odds of RT (OR 0.32; 95%CI 0.11–0.93) (Table 4). A multivariable analysis could not be performed for this subgroup due to the low number of events.

Table 4:

Univariate logistic regression: Predictors of residual tumor among clinical complete responders (cCRs)

Variables	OR	95% CI	p-value
Induction TNT	1.61	0.64–4.09	0.3
cN positive	2.45	0.88–7.94	0.087
Flat scar	0.32	0.11–0.93	0.037
Telangiectasia	0.95	0.37–2.6	>0.9

Univariable analysis of predictors of residual tumor among clinical complete responders. Bolded numbers are statistically significant. Abbreviations: OR = odds ratio, 95% CI= 95% confidence interval, TNT= total neoadjuvant therapy, cN positive = clinical node status positive

Endoscopic Features Among Near Complete Clinical Responders

Of the 99 patients with a nCR, the most common endoscopic features at restaging were nodularity (43.4%), irregular mucosa (39.4%) and ulcer (29.3%). A substantial portion (25.2%) also exhibited characteristics of a complete responder, including 19 (19.2%) with a flat scar and 17 (17.2%) with telangiectasia. The majority of nCRs (59.6%) had residual tumor, including 38.4% who developed local tumor regrowth. Compared to the NRT group, patients with RT did not differ across demographics, treatment arm or tumor characteristics. In a backwards-selected multivariable logistic regression model accounting for treatment arm, ulcer, nodularity and irregular mucosa; ulcer (OR 4.71, 95% CI 1.63–15.8) and irregular mucosa (OR 2.65, 95% CI 1.06–7.02) remained independent predictors of RT (Table 5).

Table 5:

Multivariable logistic regression: Predictors of residual tumor among near complete responders (nCRs)

Variables	OR	95% CI	p-value
Induction TNT	2.25	0.94–5.54	0.068
Ulcer	4.71	1.63–15.8	0.003
Nodularity	1.96	0.8–5.05	0.144
Irregular mucosa	2.65	1.06–7.02	0.036

Multivariable logistic regression analysis of predictors of residual tumor among near complete responders. Variables for the model were chosen using backwards selection. Bolded numbers are statistically significant. Abbreviations: OR = odds ratio, 95% CI= 95% confidence interval, TNT= total neoadjuvant therapy

DISCUSSION

Patients with LARC treated with TNT who have visible tumor, ulcer, nodularity or irregular mucosa on restaging endoscopy have higher odds of harboring RT. This is the first study using a prospectively collected and standardized endoscopic grading system to evaluate the implications of specific endoscopic features. These findings provide important clinical context for providers, as characteristics with negative prognostic value place patients at higher risk for failing NOM. The results of this study were originally presented at the annual meeting of the American Society of Colon and Rectal Surgeons.²²

Early experience with WW restricted eligibility for NOM to patients with a cCR at restaging. Habr-Gama et al. first standardized the endoscopic definition of a cCR in 2010 to include those with whitening of the mucosa, telangiectasia, and no positive findings of residual cancer.¹⁹ According to these criteria, any patients with mucosal abnormalities underwent surgery. Multiple studies have since shown that a substantial number of patients who did not meet the strict definition of a cCR had no residual cancer on pathology.^10,16,18 Additional work by Habr-Gama et al. has demonstrated that most patients do not achieve a luminal cCR until 16 weeks after TNT, highlighting that minor mucosal abnormalities at restaging may fully resolve with time.¹³ Collectively, these findings suggest that allowing nCRs to pursue a WW strategy could greatly increase the number of patients eligible for NOM.^5,10,23 Extending the observation period for patients with a nCR appears to have no adverse effect on rates of local disease control, as almost all are salvageable with TME.^24–27

There is no international consensus regarding which features delineate a nCR. The OPRA trial proposed the first standardized endoscopic definition, including ulceration, nodularity, mild erythema of the scar and mucosal irregularity (Supplementary Figure 2 at https://links.lww.com/DCR/CXX).⁴ Mild erythema of the scar appears to hold little prognostic value, as the feature was not associated with either presence or absence of tumor. Of the remaining characteristics, ulceration was associated with the highest odds of RT (OR 6.66), followed by irregular mucosa (OR 3.66) and nodularity (OR 2.96). Although nodularity was predictive of RT when evaluating the entire cohort, this association disappeared on a subgroup analysis of patients classified as nCRs by TAF criteria. This discrepancy most likely arises from variation in TAF responses. While most surgeons selected “nodularity” as a characteristic of a nCR, others selected this trait as a descriptor of visible disease seen in an iCR. After excluding patients classified as iCRs, this feature no longer had an independent association with RT. These results demonstrate that patients with ulcer or irregular mucosa on restaging endoscopy have an increased likelihood of harboring residual disease. Should these patients choose to pursue a WW strategy, they may benefit from a short interval reassessment. The prognostic implications of nodularity on restaging endoscopy remain less clear, and defining this characteristic’s risk will likely require larger sample sizes.

We did not find any endoscopic characteristics associated with lower odds of RT when analyzing the entire cohort. However, on a subgroup analysis of cCRs, patients with a flat scar had 68% lower odds of RT compared to patients without this feature. This result is consistent with a study by Van der Sande et al., which found that a flat scar was associated with a complete response.¹¹ While our cohort only contained a few patients with mixed characteristics of a cCR and a nCR, future research should investigate whether presence of a flat scar is associated with achieving a sustained cCR in this group.

Patients with clinical node positivity and those that received INCT-CRT had higher odds of RT on multivariable analysis. These findings are consistent with previously published data. The OPRA trial demonstrated that baseline node status was independently associated with disease-free survival and time to total mesorectal excision.⁴ Although studies have not shown a difference in oncologic outcomes based on TNT regimen, both the OPRA and the CAO/ARO/AIO-12 trials demonstrated that patients who received CRT-CNCT instead of INCT-CRT had higher rates of cCR and pathologic complete response, respectively.^4,28

This study has several limitations. First, this analysis did not incorporate information from subsequent endoscopic assessments. Measuring the persistence or disappearance of certain characteristics over time, particularly for nCRs, may provide additional prognostic information regarding features associated with failure of NOM. Second, our models did not take into account findings from restaging MRIs. In practice, these results influence clinical decision making and increase the accuracy of categorizing tumor response.¹¹ Third, endoscopic exams and their interpretation are inherently subjective. However, this limitation exists for any method of clinical assessment used as a surrogate marker of pathologic complete response. Fourth, the design of the TAFs created several inconsistencies that were difficult to interpret. The forms did not require surgeons to specify the absence of certain characteristics on exam. As a result, we could only comment on the presence of a feature (i.e., ulcer), but could not determine whether other patients specifically lacked that characteristic (i.e., no ulcer). This ambiguity limited our ability to evaluate multiple predictive features for cCRs. When grading tumor response by DRE, the TAF did not distinguish between a normal exam and a non-palpable proximal tumor. This, along with the substantial number of patients who did not have a DRE recorded, limited our ability to incorporate these findings into a predictive model. Finally, the patient’s surgeon presumably knew the results of the DRE before performing the flexible sigmoidoscopy. While unlikely, it is possible that this information could have influenced the endoscopic features selected.

CONCLUSION

Using prospectively collected and standardized data, this study demonstrated that several endoscopic features of a nCR are independently associated with residual disease. While extending the observation period for nCRs appears to carry minimal immediate risk, presence of ulceration or irregular mucosa on restaging endoscopy should inform providers that these patients have an increased likelihood of failing NOM. Patients with high-risk characteristics who pursue WW may require shorter interval follow up to monitor for residual disease.

COLLABORATORS

OPRA Consortium

Sujata Patil, Ph.D.; Marc J. Gollub M.D.; Jin K. Kim, M.D.; Jonathan B. Yuval, M.D.; Meghan Lee, B.S.; Richard F. Dunne, M.D.; Jorge Marcet, M.D.; Peter Cataldo, M.D.; Blase Polite, M.D.; Daniel O. Herzig, M.D.; David Liska, M.D.; Samuel Oommen, M.D.; Charles M. Friel, M.D.; Charles Ternent, M.D.; Andrew L. Coveler, M.D.; Steven Hunt MD; Anita Gregory MD; Madhulika G Varma MD; Brian L Bello MD; Joseph C Carmichael, M.D.; John Krauss, M.D.; Ana Gleisner, M.D.; Philip B. Paty, M.D.; Martin R. Weiser, M.D.; Garrett M. Nash, M.D.; Emmanouil Pappou, M.D.; José G. Guillem, M.D.; Larissa Temple, M.D.; Iris H. Wei, M.D.; Maria Widmar, M.D.; Neil H. Segal, M.D., Ph.D.; Andrea Cercek, M.D.; Rona Yaeger, M.D.; J. Joshua Smith, M.D., Ph.D.; Karyn A. Goodman; Abraham J. Wu, M.D.; and Leonard B. Saltz, M.D.