Introduction

Colorectal cancer ranks as the third most prevalent cancer globally and is the second leading cause of cancer-related deaths [1]. Rectal cancer constitutes approximately 30% of newly diagnosed colorectal cancer cases, reaching an estimated incidence of 732,210 worldwide in 2020 [1,2,3]. The conventional treatment for locally advanced rectal cancer (LARC) involves a combination of neoadjuvant chemoradiotherapy, followed by total mesorectal excision (TME) as the standard surgical procedure for rectal cancer [4], along with adjuvant chemotherapy [5]. However, patients undergoing total TME may experience long-term complications affecting bowel, urinary, and sexual function, potentially requiring a temporary or permanent ostomy [4].

The advent of the nonsurgical “watch-and-wait” approach has marked a paradigm shift in rectal cancer management. This approach prioritizes a balance between enhancing oncological outcomes and minimizing functional disabilities resulting from curative-intent surgery [6]. Comparative studies have revealed that patients opting for the watch-and-wait strategy enjoy a greater quality of life, improved function-related outcomes, reduced overall costs, and comparable cancer-specific survival compared to those undergoing TME [7,8,9,10,11]. However, the watch-and-wait strategy is associated with a higher local recurrence rate [11].

Following neoadjuvant therapy, up to one-third of patients achieve a pathological complete response (pCR), characterized by the absence of viable tumor cells in the surgically resected specimen (T0N0). This subset of patients exhibits more favorable long-term outcomes in terms of recurrence and survival than their non-pCR counterparts [12,13,14]. The complete clinical response (CCR) criteria, initially described by Habr-Gama et al., involve clinical, endoscopic, and radiologic assessments post-neoadjuvant treatment [15]. Patients meeting the CCR criteria are primarily managed nonoperatively. However, these criteria have shown insufficient accuracy in predicting pCR in various studies, lacking both sensitivity and specificity [12, 16,17,18,19,20].

Given the limitations of the CCR criteria, numerous studies have sought to identify potential predictors of pCR, particularly those ascertainable before surgery. Factors such as gross tumor volume (GTV) [21], size [22, 23], tumor circumference [24], tumor histopathology [25], grade [26], tumor distance from the anal verge [22], endoscopic findings [24], clinical T and N staging [25, 26], the neoadjuvant treatment-surgery interval [22, 26], the serum carcinoembryonic antigen (CEA) level [23, 24, 27, 28], and the C-reactive protein (CRP) level [29] have been associated with tumor response to neoadjuvant treatment and have been identified as predictors of pCR.

This study aimed to comprehensively investigate the role of laboratory and endoscopic findings in predicting pCR in patients with LARC. Additionally, we report the accuracy of each endoscopic feature in predicting pCR in patients who have undergone neoadjuvant treatment.

Materials and methods

Study population

This retrospective cohort study was performed in the Department of Colorectal Surgery, Imam Khomeini Hospital Complex, Tehran University of Medical Sciences, between April 2018 and March 2020. All consecutive patients who received neoadjuvant treatment for LARC were identified. The records of patients who met four criteria were collected: (1) diagnosed with locally advanced (T3 or T4) primary rectal cancer without distant metastasis (M0), based on the tumor characteristics in magnetic resonance imaging (MRI), computed tomography (CT) scans, and flexible proctosigmoidoscopy performed by experienced colorectal surgeons at the center, (2) received neoadjuvant treatment based on National Comprehensive Cancer Network (NCCN) guidelines, (3) underwent restaging flexible proctosigmoidoscopy within 6–8 weeks after the end of neoadjuvant treatment, and (4) underwent TME surgery within an interval of 8–12 weeks.

Demographic and clinical information, including age, sex, pre-neoadjuvant and postsurgery tumor pathological differentiation, pre-neoadjuvant and presurgery staging, location of the primary tumor, type of surgery, body mass index (BMI), the time interval between chemoradiotherapy and surgery in the neoadjuvant setting, proctosigmoidoscopic view, T stage, N stage, pathologic findings in addition to the preoperative polymorphonuclear (PMN) and lymphocyte cell percentage, hemoglobin, erythrocyte sedimentation rate (ESR), CRP, albumin, and metastases, the tumor’s distance from the anal verge (from the flexible proctosigmoidoscopy examination) before neoadjuvant chemoradiotherapy, and changes in serum CEA levels before and after neoadjuvant treatment, were extracted from patients' initial records and entered into a checklist.

The study was approved by the Iman Khomeini Hospital Complex Committee for Ethics in Biomedical Research (IR.TUMS.IKHC.REC.1398.182). This manuscript is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for the reporting of observational studies [30].

Endoscopy-based image analysis

For each restaging endoscopic (flexible proctosigmoidoscopy) image, four colorectal surgeons with at least 5 years of experience from the colorectal department of the Imam Khomeini Hospital Complex in Tehran observed and analyzed the proctosigmoidoscopy images. All photos were double-checked and independently examined by two surgeons not involved with the initial surgeon. Any disagreements were resolved through discussion. The surgeons did not have access to pre-neoadjuvant therapy colonoscopy images of the tumor. The surgeons determined the possible lesions according to any of the following endoscopic features: flattening of the marginal tumor, regeneration of the epithelium covering the central ulcer, a flat whitish scar ulcer surrounded by normal mucosa, small ulcers with regular edges, a deep ulcer, a raised ulcer, disappearance of the neoplastic pit pattern, the disappearance of the neoplastic nodule or stenosis, a passed stenosis scope, a missed stenosis scope, telangiectasia, a flat ulcer, a regular ulcer edge, an irregular ulcer edge, a pedunculated polyp, a sessile polyp, a circumferential mass, an ulcerated mass, or a nonulcerated mass.

Determination of the pathological complete response

Surgical resection of specimens is the gold standard for histopathological staging. According to Quirke and Dixon’s technique, pathologists at Imam Khomeini Hospital Complex used the TNM staging classification to define the surgical specimens’ response to treatment [31]. pCR was defined as T0N0 when no tumor cells could be found in the histological examination of the TME specimen; ypT1–4 with any N or ypN > 0 with any T stage was defined as having a residual tumor. The conventional 5-point tumor regression grade (TRG) (grades 0–4) is indicative of therapeutic response in rectal cancer patients following chemoradiotherapy preceded by curative resection [32].

Statistical analysis

The statistical analyses were performed using IBM® SPSS Statistics 22.0 (IBM® Corporation, Armonk, NY), with the presentation of categorical data in numbers (frequencies) and continuous data as the means ± standard deviations. To compare nominal data frequencies between study groups, chi-square, or Fisher’s exact tests were employed. Additionally, the independent t test was applied to compare the mean values of continuous data between groups. A p value < 0.05 was considered to indicate statistical significance for all analyses.

For predictive modeling, a univariate binary logistic regression analysis was conducted to assess the predictive power of each endoscopic feature in predicting treatment response (pCR) in patients, distinguishing between complete and non-complete response groups. The sensitivity, specificity, accuracy, odds ratio (OR) with a 95% confidence interval (CI), and p value were calculated for each variable to evaluate its predictive ability. Subsequently, a multivariate binary logistic regression analysis was performed to assess the independent predictive ability of each endoscopic variable that successfully predicted treatment response in the univariate analysis, with related p values and odds ratios with 95% CIs. The significance level for these analyses was set at a p value < 0.05.

In the evaluation of predictive values, receiver operating characteristic (ROC) curves were generated for ESR and CEA to detect pCR. The area under the ROC curve (AUC) and associated p values were calculated. Additionally, Youden’s index (sensitivity + specificity—1) was utilized to determine the optimal cutoff values for the ESR and CEA, along with their corresponding sensitivity and specificity values.

Results

Participants’ characteristics

A total of 305 patients diagnosed with primary rectal cancer and receiving neoadjuvant chemoradiotherapy underwent surgery at the designated medical center from April 2018 to March 2020. Of these, 32 patients had a T stage < 3, 11 had distant metastasis, and 143 had unavailable proctosigmoidoscopy images, and thus were excluded from the analysis. Consequently, the present study focused on a cohort of 119 patients with LARC, among whom 21 (17.6%) individuals met the criteria for pCR and were classified as the complete response group, and 119 (82.4%) were classified as the noncomplete response group.

The mean ages of the complete and non-complete response groups were 55.0 ± 13.3 and 57.1 ± 13.1 years, respectively, with no significant difference (p value = 0.500). Additionally, there were no significant differences between the two groups in terms of tumor pathology (p value = 0.114), tumor location (p value = 0.920), and tumor distance from the anal verge (p value = 0.612). Furthermore, no significant differences were observed between the groups before neoadjuvant therapy in their T and N stages of tumors (p value = 0.136 and p value = 0.733, respectively).

Regarding laboratory findings, the CEA levels before neoadjuvant therapy (2.4 ± 1.9 vs. 6.8 ± 6.9; p value = 0.010) and before surgery (2.1 ± 1.8 vs. 4.5 ± 3.1; p value < 0.001), as well as ESR (13.7 ± 8.6 vs. 28.8 ± 22.1; p value < 0.001) and CRP (13.3 ± 10.9 vs. 29.9 ± 37.5; p value = 0.009), were significantly lower in the complete response group compared to the non-complete response group. However, there were no significant differences between the groups in terms of pre-surgery WBC levels (p = 0.370), PMN ratios (p value = 0.307), and lymphocyte ratios (p value = 0.413). Detailed demographic, clinical, and endoscopic findings of the two study groups are provided in Table 1.

Table 1 Demographic, clinical, and surgical information of rectal cancer patients within the noncomplete and complete response groupsa
Full size table

Flexible proctosigmoidoscopy findings

The prevalence of telangiectasia (47.6% vs. 3.1%; p value < 0.001), flat whitish scar surrounded by normal mucosa (47.6% vs. 3.1%; p value < 0.001), flat ulcer (14.3% vs. 2.0%; p value = 0.038), flattening of marginal tumor swelling (76.2% vs. 3.1%; p value < 0.001), disappearance of the neoplastic pit pattern (33.3% vs. 7.5%; p value = 0.028), disappearance of the neoplastic nodule or stenosis (65.0% vs. 1.4%; p value < 0.001), and regenerating of epithelium covering the central ulcer (80.0% vs. 14.3%, p value < 0.001) were significantly greater in the complete response group. In contrast, circumferential mass (9.5% vs. 33.7%; p value = 0.034) and ulcerated mass (9.5% vs. 50.0%, p value = 0.001) were more common in the noncomplete response group.

Notably, no significant differences were detected between the study groups in terms of the prevalence of deep ulcers, raised ulcers, ulcer edges (regular vs. irregular), pedunculated and sessile polyps, nonulcerated masses, stenosis-scope passed, or stenosis-scope not passed across the two study groups (p values > 0.05) (Table 2).

Table 2 Comparison of frequency of endoscopic features between complete response and noncomplete response groupsa
Full size table

Prediction of treatment response with endoscopic findings

The predictive capacity and independence of endoscopic findings in relation to treatment response were investigated through both univariate and multivariate logistic regression analyses, as detailed in Table 3. Notably, the variables “disappearance of the neoplastic pit pattern” and “disappearance of the neoplastic nodule of stenosis” were excluded from the multivariate logistic regression analysis due to their considerable missing data.

Table 3 Results of univariate and multivariate analysis to evaluate the strength and independence of endoscopic findings in predicting response to treatment
Full size table

Univariate analyses demonstrated that the presence of telangiectasia (p value < 0.001), a flat whitish scar surrounded by a normal mucosa (p value < 0.001), a flat ulcer (p value = 0.028), flattening of marginal tumor swelling (p value < 0.001), the disappearance of the neoplastic pit pattern (p value = 0.025), the disappearance of the neoplastic nodule or stenosis (p value < 0.001), and regeneration of the epithelium covering the central ulcer (p value < 0.001) successfully predicted a better treatment response in patients. On the other hand, the presence of a circumferential mass (p value = 0.042) and ulcerated mass (p value = 0.003) predicted a worse treatment response. The other endoscopic variables failed to successfully predict the treatment outcomes (p values > 0.05).

Furthermore, in the multivariate analysis, only the flattening of marginal tumor swelling (p value < 0.001) predicted a better treatment response, with a calculated odds ratio of 100.605 (95% CI: 10.849–932.960). Detailed information regarding the measured p value, sensitivity, specificity, accuracy, and odds ratio with 95% CI for each variable in predicting pCR is provided in Table 3.

ROC analysis of the utility of CEA and ESR in predicting pCR

ROC analysis was conducted to assess the ability of preoperative CEA and ESR levels to predict pCR. The preoperative CEA demonstrated an AUC of 0.771, with an optimal preoperative CEA cut-off value of 2.15 ng/ml. At this cut-off, the sensitivity for detecting complete responses was 72.9%, and the specificity was 71.4% (Fig. 1). Furthermore, the AUC for the preoperative ESR was 0.739, with a measured optimal preoperative ESR cutoff of 19.0 mm/h. At this cutoff, treatment response was predicted with a sensitivity and specificity of 67.5% and 76.2%, respectively (Fig. 2).

Fig. 1
figure 1

The power of preoperative CEA in predicting surgical response

Full size image
Fig. 2
figure 2

The power of preoperative ESR in predicting surgical response

Full size image

Discussion

Our comprehensive analysis revealed that presurgery endoscopic features, along with CEA and ESR, serve as significant predictors of achieving pCR in patients with LARC. These findings underscore the potential of leveraging the endoscopic and biochemical profiles of rectal cancer patients to inform and guide the selection of optimal treatment management strategies.

The careful identification of suitable candidates for nonsurgical and conservative strategies requires the establishment of comprehensive diagnostic clinical criteria. The initial criteria proposed by Habr-Gama et al. for the adoption of the wait-and-watch strategy appear to be stringent [15]. This method selectively opts for patients exhibiting a flat scar and no mucosal defects, emphasizing organ preservation [15]. However, this approach may inadvertently exclude individuals who have a satisfactory or complete pathological response and who could benefit from organ preservation. Many such patients end up undergoing radical surgery, foregoing the opportunity to retain the affected organ. Expanding the selection criteria for organ preservation holds the promise of identifying a broader pool of patients with a complete pathological response. While this approach could lead to an increased number of patients experiencing regrowth or harboring residual tumors during follow-up, it also provides the opportunity for well-responding patients with malignancies to undergo further regression. Importantly, broadening the criteria for patients opting for an organ preservation approach provides additional time for responsive malignancies to continue regressing, affording physicians more time and precision in formulating an appropriate treatment plan [33].

Habr-Gama et al. established that certain endoscopic characteristics, such as mucosal whitening, telangiectasia, and the absence of ulceration, nodules, or stenosis, were highly indicative of achieving a complete pathological response [34]. These criteria have been widely utilized in various clinical trials focused on organ preservation to evaluate a patient’s complete clinical response [34, 35]. However, Nahas et al. reported that only 27% of patients who achieved a complete pathological response met the criteria set by Habr-Gama during restaging [17]. In a primary investigation, Hiotis et al. employed proctoscopy to assess therapeutic response, yielding disappointing results with poor accuracy. This approach detected only 50% of patients with pathological complete responses, and 25% of those with complete responses still exhibited persistent lesions [19]. Another study analyzing endoscopic images to predict a complete pathological response reported a sensitivity ranging from 69 to 87%, while the specificity varied from 39 to 74% [36]. Maas et al. demonstrated that clinical examination using endoscopy and digital rectal examination (DRE) outperformed MRI in identifying complete pathological responses, with a sensitivity of 53% and specificity of 97% [37]. Similarly, an investigation by Van der Sander et al. revealed that endoscopy had a sensitivity of 72–94%, specificity of 61–85%, positive predictive value of 63–78%, and negative predictive value of 80–89% for identifying complete clinical responses. The presence of a flat scar emerged as the most significant indicator of a complete response, with a positive predictive value of 70–80% [38]. In our study, only two endoscopic findings, namely, telangiectasia (87% accuracy) and flattening of marginal tumor swelling (91.5% accuracy), independently predicted pCR, while other features lacked predictive ability.

While early studies have explored the endoscopic features of post-neoadjuvant rectal cancer and their correlation with a pathological complete response, our study stands out for its uniqueness in examining the most comprehensive set of endoscopic features. This includes a meticulous consideration of all influential variables and confounders in the analysis, coupled with an appropriate statistical population. Currently, there is a lack of consensus regarding the use of endoscopic criteria for predicting pCR in rectal cancer patients. Notably, the criteria for selecting patients for the watch-and-wait technique are quite stringent, underscoring the need for more extensive, population-based, and prospective research in this domain. Our study contributes to this ongoing discourse by providing a thorough investigation of the various endoscopic features, addressing the complexities of influential factors, and paving the way for a more nuanced understanding of predicting pCR in rectal cancer patients.

Our analysis investigating the predictability of pCR based on the pretreatment biochemical profile of patients with rectal cancer revealed a noteworthy association between lower presurgical CEA levels and the ESR and improved outcomes. Specifically, our findings align with prior research emphasizing a more favorable treatment response in rectal cancer patients with lower presurgical CEA levels [28, 39]. A similar study conducted in Iran identified a presurgical CEA level cutoff of 2.6 ng/mL for predicting a better treatment response in patients [40]. In our study, we identified a more conservative CEA cutoff of 2.15 ng/mL in this context, further refining predictions for treatment response. Unlike CEA, whose role in predicting treatment response in rectal cancer patients has been studied extensively, the assessment of the baseline ESR for this purpose is relatively novel. While other inflammatory markers, such as the CRP/albumin ratio, neutrophil/lymphocyte ratio, and platelet/lymphocyte ratio, have been explored previously and demonstrated to play crucial roles in predicting treatment response, data regarding the utility of the ESR in this context are lacking [41]. This underscores the necessity for future research to explore the potential utility of the ESR in predicting treatment response in rectal cancer patients, further expanding our understanding of the biochemical markers involved in treatment outcomes.

While our study specifically assessed the predictive capacity of endoscopy and laboratory findings for achieving pCR, it is important to note that in clinical practice, the evaluation of therapy response and decision-making involves the integration of information from various sources. This typically includes combining endoscopy with DRE and imaging modalities such as MRI [37, 42]. Current recommendations for postchemotherapy and radiation advocate for intraoperative MRI evaluation for all patients, regardless of whether surgery has been performed [4, 43]. Previous studies have explored diverse MRI criteria and the utilization of artificial intelligence (AI) in predicting the complete pathological response of rectal cancer patients [44,45,46]. Decision-making criteria may also encompass factors such as the distance of the tumor from the anal verge and the patient’s input. While our study primarily focused on assessing the predictive value of endoscopy and laboratory findings, it is crucial to recognize the complementary role of MRI and DRE results. A previous study demonstrated that the amalgamation of these diagnostic modalities enhances diagnostic accuracy [37]. Moreover, advanced endoscopic techniques, such as narrow-spectrum and immunofluorescence imaging, offer the potential to enable physicians to observe structures with greater precision [47]. Therefore, the results of our study should be considered in conjunction with those obtained from MRI, DRE, and potentially advanced endoscopic methods, forming a comprehensive approach to enhance diagnostic power and refine medical decision-making in the context of rectal cancer treatment.

Although our study yielded promising findings regarding the potential utility of endoscopic and biochemical variables in predicting pCR in rectal cancer patients, it is essential to interpret these results with caution due to certain limitations. First, our sample size was relatively small, which can reduce the generalizability of our findings. Second, the documentation of regression was not sufficiently recorded in the pathology records, making its evaluation for classification challenging. As a result, patients were categorized into two groups based on the pathology report provided by the pathology department: complete response (including cases where tumor cells were not identified during histological examination, T0N0) and noncomplete response, which were analyzed as dependent variables. Furthermore, the lack of biopsies collected during routine follow-up after neoadjuvant chemoradiotherapy at the center where the study was conducted prevented the inclusion of this variable as one of the endoscopic findings. Integrating endoscopic findings with DRE and MRI investigations could enhance the accuracy of patient selection for the watch-and-wait approach. Another limitation involves the absence of an assessment of disease-free survival (DFS) and overall survival (OS) in the pathological complete response group and the lack of comparisons between measured variables in these groups concerning DFS and OS. Additionally, due to the long duration of patient selection and the use of various chemoradiotherapy regimens, we failed to collect data on this aspect. Therefore, future studies are needed with more detailed pathological data and longer follow-up periods, specifically focusing on DFS and OS, and adjusting their analyses for the different chemotherapy regimens. This will provide additional insights into the effectiveness of the predictive variables identified in our study. These considerations highlight the need for ongoing research and further refinement of methodologies to enhance the robustness and applicability of the findings.

Conclusion

The findings of this study suggest that endoscopic observations can potentially aid in predicting treatment response following neoadjuvant chemoradiotherapy in patients with LARC. Specifically, the flattening of marginal tumor swelling has emerged as an independent predictor of pCR. Additionally, we have identified a lower pre-surgery ESR level as a potential predictor of treatment response, alongside the previously reported CEA levels. These findings suggest that endoscopy and laboratory results can potentially be useful as adjunct tools to other established methods, such as MRI and DRE, for selecting patients with LARC for the watch-and-wait protocol. However, our result should be interpreted with caution due to significant limitations. Despite this, the information presented can pave the way for further research aimed at developing a more reliable predictor profile for treatment response in patients with rectal cancer. We hope these findings will guide physicians and researchers in developing more personalized treatment strategies, allowing for a more conservative approach to appropriately select individuals based on their unique clinical characteristics.