Results

Patient cohort

A total of 100 consecutive patients with biopsy-proven rectal adenocarcinoma who underwent baseline and post-treatment rectal MRI after total neoadjuvant therapy were included in the final analysis. The patient selection process is shown in Figure 1. The mean patient age was 56.8 ± 11.4 years. The cohort included 64 (64.0%) men and 36 (36.0%) women. Tumors were located in the lower rectum in 38 (38.0%) patients, mid rectum in 44 (44.0%) patients, and upper rectum in 18 (18.0%) patients. On baseline MRI, 8 (8.0%) patients had cT2 disease, 72 (72.0%) patients had cT3 disease, and 20 (20.0%) patients had cT4 disease. Baseline nodal positivity was present in 78 (78.0%) patients, while 22 (22.0%) patients were node negative. The baseline demographic and imaging characteristics are summarized in Table 3.

Variable | Category/Finding | Value
Total patients | Final study cohort | 100
Age | Mean ± standard deviation, years | 56.8 ± 11.4
Sex | Male | 64 (64.0%)
Female | 36 (36.0%)
Tumor location | Lower rectum | 38 (38.0%)
Mid rectum | 44 (44.0%)
Upper rectum | 18 (18.0%)
Baseline clinical T stage | cT2 | 8 (8.0%)
cT3 | 72 (72.0%)
cT4 | 20 (20.0%)
Baseline nodal status | Node-positive disease | 78 (78.0%)
Node-negative disease | 22 (22.0%)

Post-treatment magnetic resonance imaging response

Structured post-treatment MRI assessment was successfully performed in all 100 patients using the predefined response parameters described in Table 1 and the response categorization framework shown in Table 2. Complete or near-complete imaging response was identified in 32 (32.0%) patients, incomplete response in 58 (58.0%) patients, and indeterminate response in 10 (10.0%) patients.

Among the 32 (32.0%) patients with complete or near-complete imaging response, the treated tumor bed showed predominant low T2 signal fibrosis, marked regression of tumor bulk, absence of definite focal diffusion restriction, and no suspicious residual extramural vascular invasion or threatened mesorectal fascia. Representative complete or near-complete response patterns are shown in Figures 2A-2D.

Among the 58 (58.0%) patients with incomplete imaging response, persistent intermediate T2 signal at the treated tumor bed was observed in 50 (86.2%) patients, focal diffusion restriction in 47 (81.0%) patients, residual extramural vascular invasion in 22 (37.9%) patients, suspicious residual mesorectal or lateral pelvic lymph nodes in 31 (53.4%) patients, and threatened mesorectal fascia in 18 (31.0%) patients. Representative incomplete response and adverse post-treatment imaging features are shown in Figures 2E-2L.

Definitive management and histopathological outcomes

Of the 100 patients, 78 (78.0%) patients underwent total mesorectal excision, 6 (6.0%) patients underwent local excision, and 16 (16.0%) patients were selected for non-operative watch-and-wait surveillance after multidisciplinary assessment. Histopathological correlation was available in 84 surgically treated patients (84.0%). Among these 84 patients, pathological complete response was observed in 21 (25.0%) patients, while residual viable tumor was identified in 63 (75.0%) patients. Pathological nodal positivity was present in 28 (33.3%) patients. Circumferential resection margin involvement was identified in 7 (8.3%) patients, all of whom had incomplete or indeterminate response on post-treatment MRI.

In the surgical subgroup, complete or near-complete imaging response was observed in 24 (28.6%) patients. Of these, 17 (70.8%) patients had pathological complete response and 7 (29.2%) patients had residual microscopic or small-volume viable tumor. Incomplete imaging response was observed in 52 (61.9%) surgically treated patients, of whom 49 (94.2%) patients had residual viable tumor and 3 (5.8%) patients had pathological complete response. Indeterminate imaging response was observed in 8 (9.5%) surgically treated patients, of whom 7 (87.5%) patients had residual viable tumor, and 1 (12.5%) patient had pathological complete response.

Correlation of imaging findings with residual disease

Persistent intermediate T2 signal, focal diffusion restriction, residual extramural vascular invasion, suspicious residual lymph nodes, and threatened mesorectal fascia were more frequently observed in patients with residual viable tumor than in those with pathological complete response. Focal diffusion restriction was present in 47 of 63 patients (74.6%) with residual viable tumor and in 3 of 21 patients (14.3%) with pathological complete response. This association was statistically significant on Pearson chi-square testing (χ² = 23.784, p < 0.001). Persistent intermediate T2 signal was present in 52 of 63 patients (82.5%) with residual viable tumor and in 4 of 21 patients (19.0%) with pathological complete response, also showing a statistically significant association with residual viable tumor on Pearson chi-square testing (χ² = 28.571, p < 0.001).

Residual extramural vascular invasion was present in 22 of 63 patients (34.9%) with residual viable tumor and in 0 of 21 patients (0.0%) with pathological complete response. This association was statistically significant on Fisher’s exact testing (p = 0.001). Threatened mesorectal fascia was present in 18 of 63 patients (28.6%) with residual viable tumor and in 0 of 21 patients (0.0%) with pathological complete response, showing a statistically significant association on Fisher’s exact testing (p = 0.004). These findings indicate that persistent intermediate T2 signal, focal diffusion restriction, residual extramural vascular invasion, and threatened mesorectal fascia were significantly associated with residual viable tumor on histopathology.

When incomplete or indeterminate response was considered positive for significant residual disease, structured MRI showed a sensitivity of 88.9%, specificity of 81.0%, positive predictive value of 93.3%, negative predictive value of 70.8%, and overall accuracy of 86.9% for detecting residual viable tumor in the surgical cohort. The association between structured MRI response category and histopathological residual viable tumor was statistically significant on Pearson chi-square testing (χ² = 37.644, p < 0.001). The diagnostic performance of structured post-treatment MRI is summarized in Table 4, and the association between structured MRI response category and histopathological outcome is shown in Table 5.

Diagnostic parameter | Numerator/Denominator | Value (%) | 95% CI | Interpretation
Sensitivity | 56/63 | 88.9 | 78.4-95.4 | Detection of residual viable tumor
Specificity | 17/21 | 81.0 | 58.1-94.6 | Correct identification of pathological complete response
Positive predictive value | 56/60 | 93.3 | 83.8-98.2 | Probability of residual tumor when MRI was positive
Negative predictive value | 17/24 | 70.8 | 48.9-87.4 | Probability of no residual tumor when MRI was negative
Accuracy | 73/84 | 86.9 | 77.8-93.3 | Overall diagnostic agreement with histopathology

Structured MRI response category | Residual viable tumor present, n (%) | Pathological complete response, n (%) | Total, n (%)
Incomplete or indeterminate response | 56/63 (88.9%) | 4/21 (19.0%) | 60/84 (71.4%)
Complete or near-complete response | 7/63 (11.1%) | 17/21 (81.0%) | 24/84 (28.6%)
Total | 63/84 (75.0%) | 21/84 (25.0%) | 84/84 (100.0%)
Pearson chi-square test | χ² = 37.644 | p < 0.001 | Statistically significant

Exploratory predictive, receiver operating characteristic, and sensitivity analyses

Exploratory predictive analysis was performed in the 84 surgically treated patients with histopathological reference standard. Because the number of pathological complete response events was limited and several imaging variables were overlapping components of the structured MRI response category, formal multivariable logistic regression was considered but was not used to create a definitive clinical prediction model. Persistent intermediate T2 signal showed a strong association with residual viable tumor, with an odds ratio of 20.09 and a 95% confidence interval of 5.65-71.44. Focal diffusion restriction was also strongly associated with residual viable tumor, with an odds ratio of 17.63 and a 95% confidence interval of 4.58-67.82. Residual extramural vascular invasion and threatened mesorectal fascia were seen only in patients with residual viable tumor and not in patients with pathological complete response; therefore, Haldane-Anscombe correction was applied for odds ratio estimation. The corrected odds ratio was 23.31 for residual extramural vascular invasion and 17.48 for threatened mesorectal fascia. These findings support persistent intermediate T2 signal and focal diffusion restriction as the most stable imaging predictors of residual viable tumor, while residual extramural vascular invasion and threatened mesorectal fascia represent high-risk adverse post-treatment features.

Receiver operating characteristic analysis was performed using the ordered structured MRI response category as an ordinal predictor of residual viable tumor. The ordered response category showed good discriminatory performance, with an area under the curve of 0.860. Bootstrap internal validation showed an approximate 95% confidence interval of 0.759-0.944, supporting acceptable internal stability of the ordinal response-category model. When structured MRI response was dichotomized as incomplete or indeterminate response versus complete or near-complete response, the area under the curve was 0.849.

Sensitivity analysis was performed to evaluate the effect of the indeterminate MRI response category on diagnostic performance. In the primary analysis, indeterminate response was treated as positive for residual disease, giving sensitivity of 56/63 (88.9%), specificity of 17/21 (81.0%), positive predictive value of 56/60 (93.3%), negative predictive value of 17/24 (70.8%), and accuracy of 73/84 (86.9%). When indeterminate cases were excluded, sensitivity was 49/56 (87.5%), specificity was 17/20 (85.0%), positive predictive value was 49/52 (94.2%), negative predictive value was 17/24 (70.8%), and accuracy was 66/76 (86.8%). When indeterminate response was treated as negative, sensitivity decreased to 49/63 (77.8%), specificity was 18/21 (85.7%), positive predictive value was 49/52 (94.2%), negative predictive value decreased to 18/32 (56.2%), and accuracy was 67/84 (79.8%). These findings indicate that indeterminate post-treatment MRI findings should not be considered reassuring and should prompt multidisciplinary correlation, endoscopic reassessment, short-interval MRI, biopsy when appropriate, or close surveillance.

Interobserver agreement

Interobserver agreement was substantial for the overall imaging response category, with a kappa value of 0.78. Agreement was substantial for magnetic resonance tumor regression grade, with a kappa value of 0.72, and almost perfect for diffusion restriction assessment, with a kappa value of 0.81. Agreement was almost perfect for mesorectal fascia status, with a kappa value of 0.84, substantial for extramural vascular invasion status, with a kappa value of 0.76, and substantial for nodal response assessment, with a kappa value of 0.69. These interobserver agreement values were calculated from the independent reader assessments before consensus review. Because only summary kappa values were available for final manuscript analysis and the complete reader-by-reader contingency matrices were not retained in the extracted dataset, 95% confidence intervals for individual kappa values could not be reliably recalculated. This limitation was acknowledged in the statistical interpretation, and the final consensus assessment was used only after independent agreement analysis had been completed.

Watch-and-wait subgroup

In total 16 (16.0%) patients were selected for non-operative watch-and-wait surveillance after multidisciplinary review. Of these, 8 (50.0%) patients had complete or near-complete imaging response, 6 (37.5%) patients had incomplete imaging response but were managed according to individualized multidisciplinary decisions because of clinical, surgical, or patient-related factors, and 2 (12.5%) patients had indeterminate imaging findings requiring close follow-up. Watch-and-wait selection was not based on MRI alone and included multidisciplinary assessment with clinical examination, endoscopic findings, tumor marker status where available, patient fitness, surgical feasibility, and patient preference. During the available follow-up period, no definite early local regrowth was documented on follow-up MRI in the complete or near-complete response subgroup; however, follow-up duration was limited and not uniform across all patients. Therefore, the watch-and-wait findings were considered descriptive only and were not used to validate structured MRI as an independent tool for organ-preservation decision-making. Patients with incomplete or indeterminate findings underwent closer clinical, endoscopic, and imaging surveillance.