Predicting neoadjuvant chemoradiotherapy response in rectal cancer: Insights from biomarkers to clinical practice

Abstract

Rectal cancer poses a major global health challenge, with neoadjuvant chemoradiotherapy improving outcomes in locally advanced cases by reducing tumor burden and recurrence risk. However, response variability, including only 15%-20% of patients achieving pathological complete response, underscores the urgent need for accurate predictive tools. This review explored current and emerging biomarkers to enhance neoadjuvant chemoradiotherapy response prediction and inform clinical practice.

Key Words: Rectal cancer; Neoadjuvant chemoradiotherapy; Pathological complete response; Biomarkers; Tumor regression grading; Clinical predictors; Radiomics

Core Tip: This review synthesized biomarkers predicting neoadjuvant chemoradiotherapy response in rectal cancer, including clinical predictors (e.g., tumor size, nodal status), histopathological features (e.g., tumor regression grading), molecular markers (e.g., KRAS mutations), microenvironmental factors (e.g., tumor-infiltrating lymphocytes), and radiomics to guide personalized treatment and improve outcomes.

INTRODUCTION

The management of rectal cancer has evolved significantly with neoadjuvant therapies, driven by advances like total mesorectal excision (TME), which reduced local recurrence rates from up to 40% to 3.7% at 5 years[1]. Preoperative radiotherapy, established by trials like the Swedish Rectal Cancer Trial[2] and Dutch TME trial[3], further lowered recurrence (e.g., 2.4% vs 8.2% at 10 years with TME alone), though overall survival (OS) benefits remained elusive[4]. Chemoradiotherapy (nCRT) trials, such as the German Rectal Cancer Trial[5], confirmed the superiority of preoperative nCRT over postoperative treatment for local control (6% vs 13% at 5 years), but systemic relapse persisted. Magnetic resonance imaging-based staging, validated by the MERCURY trial[6], refined patient stratification, identifying high-risk features like cT3c/d, threatened mesorectal fascia, and extramural venous invasion for neoadjuvant therapy[7].

CURRENT STATUS OF NEOADJUVANT THERAPIES

Current indications vary, with North America favoring neoadjuvant therapy for cT1-2N+ tumors[8], while Europe reserves it for advanced cases[9]. Short-course radiotherapy (SCRT) and long-course nCRT show comparable local control, though nCRT may enhance R0 resection in mesorectal fascia-threatened cases[10]. Total neoadjuvant therapy (TNT), integrating full-dose chemotherapy preoperatively, as in the RAPIDO trial[11], reduced distant metastases (20% vs 27%), boosting pathological complete response (pCR) rates (28% vs 14%). Neoadjuvant chemotherapy trials like PROSPECT[12] suggest FOLFOX as a de-escalation option for intermediate-risk cases, achieving similar disease-free survival (DFS) to nCRT (80.8% vs 78.6% at 5 years). Nonoperative management post-nCRT or TNT, as in OPRA[13], preserved organs in over 50% of complete responders, maintaining oncological safety.

Future directions include immunotherapy for mismatch repair-deficient tumors, with trials like NICHE[14] showing 60% pCR rates and biomarkers like circulating tumor DNA for response prediction[15]. Gut microbiome studies also highlighted potential modulators of treatment response[16]. A systematic review and network meta-analysis by Turri et al[17] evaluated neoadjuvant treatments for locally advanced rectal cancer across 27 randomized clinical trials involving 13413 adults (median age 60 years, 67.2% male), comparing TNT protocols to standard long-course nCRT with single-agent fluoropyrimidine (L-CRT1). The primary outcome, pCR, showed TNT regimens outperforming L-CRT1. L-CRT plus consolidation chemotherapy (L-CRT + consolidation) had the highest effect [relative risk (RR): 1.96, 95% confidence interval (CI): 1.25-3.06)], followed by SCRT plus consolidation (S-RT + consolidation, RR: 1.76, 95%CI: 1.34-2.30) and induction chemotherapy plus L-CRT (RR: 1.57, 95%CI: 1.09-2.25)[17]. These TNT protocols achieved pCR rates of 18.6%-21.5% vs 14.3% for L-CRT1, with L-CRT + consolidation requiring treatment of 7.5 individuals for one additional pCR. Standard L-CRT1, recommended by guidelines[18], was less effective than TNT but outperformed neoadjuvant chemotherapy alone (RR: 0.75, 95%CI: 0.57-0.98) and SCRT with early resection (S-RTearly, RR: 0.07, 95%CI: 0.02-0.22).

TNT showed good tolerability, with S-RT + consolidation (RR: 0.90, 95%CI: 0.82-0.99) and L-CRT with duplex chemotherapy (L-CRT2, RR: 0.91, 95%CI: 0.86-0.97) slightly better than L-CRT1, though toxic effects were higher (e.g., S-RT + consolidation, RR: 2.01, 95%CI: 1.39-2.91)[17]. Surgical outcomes (e.g., R0 resection rates) were comparable across treatments, but S-RT + consolidation increased 5-year locoregional recurrence (RR: 1.65, 95%CI: 1.03-2.63) while improving 3-year DFS (RR: 1.08, 95%CI: 1.01-1.14)[11]. Induction + L-CRT also enhanced 3-year DFS (RR: 1.12, 95%CI: 1.01-1.24)[19]. The study suggests TNT protocols, particularly L-CRT + consolidation, should be first-line for maximizing pCR, supported by high-to-moderate evidence certainty[20], though long-term survival benefits require further data.

LIMITATIONS OF CURRENT RECOMMENDATIONS

While the RAPIDO trial by Bahadoer et al[11] and the meta-analysis by Turri et al[17] highlighted the short-term benefits of TNT protocols, particularly in achieving higher pCR rates, long-term survival data (e.g., 5-year OS) are not yet mature. The increased 5-year locoregional recurrence with S-RT + consolidation (RR: 1.65) raises concerns about its long-term efficacy despite improved 3-year DFS. Therefore, recommendations for TNT as a first-line regimen should be cautious, and future studies should prioritize long-term outcomes to confirm its superiority over traditional nCRT.

TUMOR REGRESSION GRADING SYSTEMS

Tumor regression grading (TRG) systems assess the histopathological response of gastrointestinal carcinomas to neoadjuvant therapy, providing insights into therapeutic efficacy and prognosis. Kim et al[21] investigated four TRG systems, American Joint Committee on Cancer, Dworak, Ryan, and a modified Dworak (mDworak), in 933 patients with rectal cancer treated with preoperative nCRT and curative resection. All systems significantly predicted recurrence-free survival (RFS) and OS (P < 0.001) with mDworak, which evaluates both primary tumor and regional lymph nodes, showing superior prognostic ability (χ²: 68.92, C-statistic: 0.6492 for RFS; 58.06, 0.6783 for OS) compared to American Joint Committee on Cancer (59.58, 0.6359; 53.95, 0.6718), Dworak (61.85, 0.6374; 55.52, 0.6711), and Ryan (59.57, 0.6356; 53.76, 0.6700), though differences were not statistically significant[21]. The mDworak system defines TRG 4 as complete regression (ypT0N0), TRG 3 as near-complete regression (< 0.5 cm foci), TRG 2 as moderate regression (> 50% fibroinflammatory changes), and TRG 1 as minimal regression (> 50% tumor mass), enhancing prediction when combined with ypStage (χ² 133.35, C-statistic 0.7248 for RFS; 97.33, 0.7482 for OS)[21].

In contrast Thies and Langer[22] reviewed TRG systems across gastrointestinal cancers, including Mandard, Becker, Dworak, and Rödel, emphasizing their histopathological basis-fibrosis vs residual tumor percentage and prognostic value. Mandard TRG 1-3 (complete to moderate regression) correlated with better DFS in esophageal cancer[23], while Becker TRG 1a-b (< 10% residual tumor) predicted improved survival in gastric cancer[24]. Dworak TRG 4 (no vital tumor cells) and Rödel TRG 4 (complete regression) also indicated favorable outcomes in rectal cancer[25,26]. Both studies highlight the prognostic significance of TRG, with complete regression (e.g., mDworak TRG 4, Becker TRG 1a) linked to better survival. Limitations like interobserver variability for mDworak showed high concordance (kappa 0.936)[21], while Mandard and Dworak had lower reproducibility (kappa 0.28-0.35)[27]. Thies and Langer[22] advocated for standardized specimen processing (e.g., embedding the entire tumor bed) and favored Becker’s percentage-based system for reproducibility (kappa 0.52). Together, these findings suggest that integrating lymph node assessment (mDworak) with primary tumor regression and adopting quantifiable criteria could optimize TRG systems for clinical use (Table 1).

Table 1 Tumor regression grading systems and prognostic value.

TRG system	Description	Prognostic value	Limitations
mDworak	Evaluates primary tumor + regional LNs; TRG 4 = ypT0N0, TRG 3 = near-complete regression	Best predictor of RFS and OS (Kim et al[21]); C-statistic: 0.6492 (RFS), 0.6783 (OS)	Requires LN assessment, less commonly used
AJCC	TRG 0-3 scale (complete response to poor response)	Predictive of survival (Kim et al[21]); C-statistic: 0.6359 (RFS), 0.6718 (OS)	Moderate reproducibility
Dworak	TRG 4 = complete regression, TRG 1 = minimal regression	Associated with survival, but lower reproducibility (Chetty et al[27])	Interobserver variability (kappa 0.28-0.35)
Ryan	TRG 0-3 scale; used in multiple cancers	Predictive of recurrence and survival (Kim et al[21])	Similar performance to AJCC and Dworak
Mandard	TRG 1-5; based on fibrosis and residual tumor	Predicts DFS in esophageal cancer (Mandard et al[23])	Lower reproducibility compared to Becker (kappa 0.28)
Becker	TRG 1a-b (< 10% residual tumor) predictive of survival	High reproducibility	Less widely used outside gastric cancer
Rödel	TRG 4 = complete regression	Favorable outcomes in rectal cancer (Rödel et al[26])	Requires histopathological expertise

mDworak: Modified Dworak; LNs: Lymph nodes; TRG: Tumor regression grading; RFS: Recurrence-free survival; OS: Overall survival; AJCC: American Joint Committee on Cancer; DFS: Disease-free survival.

CLINICAL PREDICTORS AND BIOMARKERS

Clinical predictors of pCR included smaller tumor size (< 4 cm), clinical node negativity (cN0), and well-differentiated histology, as identified by Turri et al[17] in a network meta-analysis of 27 randomized clinical trials involving 13413 patients with locally advanced rectal cancer. These clinical markers (tumor size, nodal status, and differentiation) were validated by multivariate analysis[28] and suggest that less aggressive tumor phenotypes benefit most from intensified neoadjuvant strategies, though long-term outcomes like 5-year locoregional recurrence (RR: 1.65 for S-RT + consolidation) require further exploration[11].

Li et al[29] expanded the marker landscape in their review, categorizing predictors into clinical, histopathological, molecular, and TME domains. Clinical predictors extended beyond the findings of Turri et al[17] to include tumor size (< 3 cm)[30], early clinical T stage (cT1-2)[31], distance from the anal verge (e.g., < 5 cm linked to better response in some studies)[32], and low pretreatment carcinoembryonic antigen (< 5 µg/L)[33]. Histopathologically, well-differentiated tumors[34], low tumor budding[35], and absence of lymphovascular or perineural invasion[36] predicted favorable nCRT response, while mucinous histology correlated with resistance[37].

MOLECULAR MARKERS: CLARIFYING CONTRADICTORY EVIDENCE

Molecular markers included protein expression levels (e.g., low RAD18[38], high MRP3[39], low TCF-4[40], and high Beclin 1[41]) as well as genetic alterations like KRAS mutations (resistance)[42] and TP53 status (variable predictive value)[43]. Regarding KRAS mutations, while Chow et al[42] reported an association with decreased response to nCRT, other studies have shown mixed results, with some finding no significant correlation[44]. Similarly, TP53 mutations have been linked to both resistance and sensitivity depending on the study[43,45]. A meta-analysis by Chen et al[43] found that TP53 mutations were associated with poorer response to nCRT (odds ratio: 0.57, 95%CI: 0.36-0.91), though heterogeneity among studies was noted (I² = 62%).

Epigenetic markers, such as CpG island methylator phenotype[44], and microRNAs (e.g., miR-21-5p overexpression)[46] further enriched molecular predictors. TME markers highlighted high CD8+ tumor-infiltrating lymphocytes[47], low FOXP3+ regulatory T cells[48], and low M2 macrophage density[49] as favorable, with gut microbiome diversity (e.g., enriched beneficial taxa) emerging as a novel predictor[50]. Patient-derived organoids mirrored clinical responses, offering a platform to validate these markers[51].

Jiang et al[52] integrated pretreatment magnetic resonance imaging radiomics (T2-weighting imaging and diffusion-weighted imaging) with clinical factors into a nomogram, achieving an area under the curve (AUC) of 0.99 (training) and 0.94 (validation) for TRG. Clinical predictors included low carcinoembryonic antigen (< 5 µg/L)[53], negative circumferential resection margin[6], high tumor differentiation[54], and absence of extramural venous invasion[55]. Radiomics extracted 32 features from combined T2-weighting imaging/diffusion-weighted imaging sequences, capturing tumor heterogeneity[56], and outperformed single-sequence (AUC 0.72-0.91) and clinical-only models (AUC 0.77-0.83). This comprehensive model enhanced TRG prediction (good responders: TRG 1-2; poor: TRG 3-5), supporting organ-preserving strategies[57] or intensified therapy for non-responders.

LIMITATIONS OF THE JIANG MODEL

However, the high AUC (0.99 in training) of the Jiang model may overestimate its clinical universality due to potential overfitting, as the sample size and heterogeneity of the external validation cohort were not detailed. Smaller validation cohorts or lack of diverse patient populations could limit generalizability. Future studies should include larger, multicenter validations to confirm these findings. These findings are summarized in Table 2.

Table 2 Biomarkers for predicting neoadjuvant chemoradiotherapy response.

Category	Predictors	Key findings
Clinical markers	Tumor size (< 4 cm), clinical node negativity (cN0), well-differentiated histology	Associated with higher pCR rates (Turri et al[17])
	Distance to anal verge (< 5 cm), low pretreatment CEA (< 5 µg/L)	Favorable response predictors (Peng et al[33]; Shao et al[32])
Histopathological markers	Low tumor budding, absence of LVI and PNI	Correlates with better nCRT response (Agarwal et al[36]; Rogers et al[35])
	Mucinous histology	Linked to treatment resistance (Simha et al[37])
Molecular markers	KRAS mutations (resistance), TP53 status (variable), low RAD18, high Beclin 1	Genetic alterations impact therapy sensitivity (Chow et al[42]; Zaanan et al[41])
	Epigenetics: CpG island methylation, microRNAs (e.g., miR-21-5p overexpression)	Influence gene expression and therapy resistance (Jo et al[44]; Lopes-Ramos et al[46])
Tumor microenvironment	High CD8+ TILs, low FOXP3+ Tregs, low M2 macrophage density	Favorable immune landscape (McCoy et al[48]; Shabo et al[49])
	Gut microbiome diversity	Emerging predictor of response (Yi et al[50])
Radiological biomarkers	MRI-based radiomics (T2WI, DWI), CRM, EMVI	MRI features enhance response prediction (Jiang et al[52]; Taylor et al[6])
	PET-based metabolic response	Functional assessment of tumor viability post-nCRT

CEA: Carcinoembryonic antigen; pCR: Pathological complete response; LVI: Lymphovascular; PNI: Perineural invasion; nCRT: Neoadjuvant chemoradiotherapy; TILs: Tumor-infiltrating lymphocytes; MRI: Magnetic resonance imaging; CRM: Circumferential resection margin; EMVI: Extramural vascular invasion; PET: Positron emission tomography; Tregs: Regulatory T cells; T2WI: T2-weighted imaging; DWI: Diffusion-weighted imaging.

CHALLENGES IN THE CLINICAL APPLICATION OF BIOMARKERS

While biomarkers hold promise for predicting nCRT response, their clinical application faces several challenges.

Detection standardization

Variability in biomarker assays (e.g., immunohistochemistry for tumor-infiltrating lymphocytes) and lack of standardized cutoffs limit reproducibility across laboratories[27].

Cost-effectiveness

Multiomic testing (e.g., genomics, proteomics) is expensive, potentially restricting access in resource-limited settings. A cost-effectiveness analysis by Widmar et al[58] estimated that biomarker-driven nCRT personalization could increase costs by 15%-20%, though long-term savings from reduced recurrence may offset this. Simplified models using fewer biomarkers or clinical predictors alone could enhance feasibility.

Ethical issues

Biomarker-based treatment decisions raise concerns about equity, particularly if access to testing is unequal. Additionally, false positives/negatives in prediction models could lead to overtreatment or undertreatment, necessitating robust validation.

Addressing these challenges requires international collaboration to standardize assays, develop cost-effective testing platforms, and ensure equitable access to personalized care.

CLINICAL IMPLICATIONS AND FUTURE DIRECTIONS

Despite promising advances biomarker-based prediction models have not yet been widely adopted in clinical practice. Future research should focus on: (1) Integration of multiomic data (genomics, proteomics, radiomics) for enhanced predictive accuracy; (2) Prospective validation of biomarker-driven treatment algorithms to personalize rectal cancer management; (3) Cost-effectiveness studies to assess the economic feasibility of biomarker testing, especially in resource-limited settings; and (4) Long-term survival data to confirm the benefits of intensified neoadjuvant strategies like TNT.

CONCLUSION

Predicting nCRT response in rectal cancer remains an area of active investigation. Histopathological markers, alongside molecular and imaging biomarkers, hold potential in guiding treatment decisions. Incorporating these biomarkers into routine practice could refine patient stratification, leading to improved outcomes and reduced treatment morbidity. However, challenges in standardization, cost, and ethical considerations must be addressed to ensure broad clinical applicability.

ACKNOWLEDGEMENTS

The authors thank the Ankara City Hospital for institutional support.