Longitudinal evolution of low anterior resection syndrome in ultra-low rectal cancer: A trend analysis of a propensity-matched cohort

Abstract

BACKGROUND

Low anterior resection syndrome (LARS) severely compromises patients’ quality of life after sphincter-preserving surgery for rectal cancer. Although the etiology of LARS is multifactorial, the tumor location is considered a primary determinant, as it directly governs the height of the surgical anastomosis. However, it remains unclear whether the initial functional deficit and the subsequent long-term recovery trajectory differ according to the tumor height. The longitudinal evolution of LARS in patients with ultra-low rectal cancer is not well characterized. Clarifying these distinct recovery patterns is essential for developing personalized postoperative rehabilitation and functional training protocols for patients with different tumor locations.

AIM

To compare the postoperative features and longitudinal recovery trajectories of LARS between patients with ultra-low and non-ultra-low rectal cancer.

METHODS

In this single-center prospective cohort study (June 2018 to January 2024), patients undergoing sphincter-preserving surgery were stratified into ultra-low (≤ 3 cm from the dentate line) and non-ultra-low (> 3 cm) tumor groups. LARS scores were systematically collected via follow-up at 1-, 3-, 6-, 9-, and 12-month postoperatively. After propensity score matching, A generalized estimating equation (GEE) model was used to model the longitudinal LARS data, assessing the effects of group, time, and their interaction while accounting for repeated measures.

RESULTS

The final analysis included a matched cohort of 220 patients (110 per group). At all postoperative follow-up points, the ultra-low group demonstrated a significantly higher incidence of LARS (e.g., 1 month, 75.5% vs 51.8%; 12 months, 30.0% vs 14.5%; all P < 0.05). Longitudinal GEE modeling confirmed that an ultra-low tumor location was associated with a persistently higher overall odds of LARS throughout the first year (OR = 2.858, 95%CI = 1.611-5.070; P < 0.001). A significant time effect (P < 0.001) signaled functional improvement in both cohorts. Critically, the non-significant group-by-time interaction (P = 0.900) revealed that the groups followed parallel recovery trajectories.

CONCLUSION

All patients exhibited improved gastrointestinal function over the first year, but those with ultra-low tumors had significantly worse baseline function and persistently higher LARS risk, underscoring the need for tailored management.

Key Words: Rectal cancer; Low anterior resection syndrome; Tumor location; Propensity score matching; Longitudinal studies

Core Tip: Parallel recovery trajectory: After propensity score matching, the ultra-low and non-ultra-low rectal cancer groups shared a similar rate of functional improvement throughout the first postoperative year. Persistently higher risk: Despite the similar recovery pace, the ultra-low group consistently faced a significantly higher overall risk of low anterior resection syndrome (LARS), attributable to a worse initial functional baseline. Clinical implication: These findings establish tumor height as a critical, independent determinant of LARS, underscoring the need for tailored counseling and proactive management strategies for patients with ultra-low tumors.

INTRODUCTION

Colorectal cancer is the third most common malignant tumor globally, with more than 1.9 million new cases reported in 2022, approximately one-third of which were rectal cancer[1]. The landscape of cancer care has profoundly shifted over recent decades, moving beyond a singular focus on survival toward an integrated paradigm that equally prioritizes long-term quality of life (QoL). This evolution is reflected in the refinement of multimodal strategies combining surgery, neoadjuvant therapies, and supportive care, which have collectively improved oncological outcomes for patients with rectal cancer[2,3]. Concurrently, with a growing population of long-term survivors, the management of treatment sequelae that affect daily functioning and well-being has become a critical component of comprehensive cancer care[4].

Within this context, the widespread adoption of sphincter-preserving surgery for rectal cancer epitomizes the pursuit of organ preservation and QoL[5]. However, this approach often incurs a functional cost, most notably a high incidence of postoperative bowel dysfunction[6,7]. A major challenge in achieving this goal is the high incidence of postoperative gastrointestinal dysfunction following sphincter-preserving surgery, affecting an estimated 30%-60% of patients[8,9]. This cluster of symptoms is formally known as low anterior resection syndrome (LARS), defined as “disordered bowel function after rectal resection, leading to a detriment in QoL”. LARS manifests primarily through symptoms such as abnormal stool frequency, fecal incontinence, and evacuation difficulties, all of which significantly impair patients’ QoL[10]. To standardize its assessment, the LARS score was developed in 2012 as a dedicated tool[11-15], and it has since been validated and widely adopted in numerous countries because of its strong correlation with QoL outcomes[16-19].

Current research on LARS has largely concentrated on identifying risk factors and exploring potential intervention strategies. Among a multitude of factors, the tumor location is among the most widely recognized determinants of LARS risk, as it directly dictates anastomotic height and the extent of rectal resection[20-22]. Consequently, ultra-low rectal cancers pose a significantly higher risk. However, current evidence largely provides a static, cross-sectional snapshot of risk. A critical gap is the lack of understanding regarding the longitudinal trajectory of functional recovery. It remains unclear whether patients with ultra-low tumors both start with worse function and experience a different pace or pattern of recovery than those with higher tumors.

Therefore, this study was designed with a strategic angle aligned with cancer care philosophies that look beyond malignant cells to the functional integrity of the host organ system[23]. We aimed to characterize and compare the postoperative features and longitudinal recovery trajectories of LARS between patients with ultra-low and non-ultra-low rectal cancer. By clarifying these distinct patterns, we aimed to provide a robust evidence base to inform clinical decision-making, guide tailored postoperative rehabilitation strategies, and ultimately improve the QoL of rectal cancer survivors.

MATERIALS AND METHODS

Study design and participants

This prospective cohort analysis was based on the Database of Colorectal Cancer (DACCA) at West China Hospital of Sichuan University. From this established cohort, we prospectively enrolled patients who underwent rectal cancer surgery between June 2018 and January 2024 for this specific LARS-focused investigation.

Eligible patients met the following inclusion criteria: (1) Age ≥ 18 years; (2) Pathologically confirmed rectal cancer; (3) Prior sphincter-preserving surgery; and (4) Conscious and able to understand and complete the LARS score questionnaire. Conversely, patients with pre-existing functional gastrointestinal disorders or inflammatory bowel disease and those who died or were lost to follow-up within 12 months were excluded from the study.

Data collection and variables

Standard demographic, clinical, and surgical data were systematically collected for all eligible patients from the electronic medical records of West China Hospital. The collected preoperative data included demographic information and baseline comorbidities. Intraoperative details, tumor characteristics, and early postoperative outcomes were documented during the hospitalization period.

The key exposure variable was the tumor location, which was determined on the basis of intraoperative findings. Patients were stratified into the ultra-low (≤ 3 cm) or non-ultra-low group (> 3 cm) based on the distance from the distal tumor edge to the dentate line (Figure 1). The covariates collected for analysis included age, sex, body mass index (BMI), blood type, hypertension, diabetes mellitus, tumor histology, TNM stage, neoadjuvant therapy (chemotherapy and radiotherapy), and postoperative complications.

Figure 1 Schematic illustration of tumor location grouping based on distance from the dentate line. Rectum icon by Servier (https://smart.servier.com/) is licensed under CC-BY 3.0 Unported (Supplementary material).

The primary outcome was the LARS score, an internationally validated, symptom-based scoring system for evaluating postoperative bowel function[11]. The score comprises five items: Incontinence for flatus, incontinence for liquid stool, frequency of bowel movements, clustering of stools, and urgency. The total score ranges from 0 to 42, classifying patients into three severity groups: No LARS (0-20 points), minor LARS (21-29 points), and major LARS (30-42 points). For this study, the presence of LARS was defined as a score of 21 or higher. We utilized the validated Chinese version of the LARS questionnaire for all assessments[14].

Postoperative follow-up was conducted by professionally trained team members at 1-, 3-, 6-, 9-, and 12-month. Patients were considered lost to follow-up and excluded from the final analysis if they died or if they were otherwise unable to complete the scheduled LARS follow-up assessments. Furthermore, a patient was defined as lost to follow-up if he or she could not be successfully contacted for data acquisition after three separate attempts by both telephone and online messaging over a continuous 2-week period.

Statistical analysis

All statistical analyses were performed using SPSS version 26.0 (IBM, Armonk, NY, United States), with a two-sided significance level set at α = 0.05.

Baseline demographic and clinical characteristics were presented as the mean ± SD for continuous variables and as n (%) for categorical variables. Initial comparisons between the ultra-low and non-ultra-low groups were conducted using the independent-samples t-test or χ² test as appropriate. To mitigate selection bias and control for potential confounding factors from baseline differences, 1:1 propensity score matching (PSM) was performed using a nearest-neighbor matching algorithm with a caliper width of 0.2 of the standard deviation of the logit of the propensity score.

After matching, the incidence of LARS at each of the five follow-up assessments (1-, 3-, 6-, 9-, and 12-month) was compared between the two groups using the χ² test.

The primary analysis focused on the longitudinal evolution of LARS. A generalized estimating equation (GEE) model was employed to analyze the repeated binary outcome of the presence or absence of LARS at the five time points. This model was chosen for its ability to account for the intrapatient correlation of repeated measurements. The GEE model included main effects for group (ultra-low vs non-ultra-low) and time (as a categorical variable), as well as a group-by-time interaction term. An exchangeable working correlation matrix was specified. This approach allowed us to assess the overall difference in LARS risk between the groups, evaluate the overall trend of functional recovery over the first year, and determine whether the rate of this recovery trend differed between the two groups. Results were presented as ORs with 95%CIs.

As a supplementary analysis, separate multivariable binary logistic regression models were constructed for each time point to further assess the association between the tumor location and LARS risk while adjusting for other relevant covariates.

Ethical considerations

This research was registered with the Chinese Clinical Trial Registry (registration No. ChiCTR2100048467).

This study was conducted in strict accordance with the Declaration of Helsinki and was approved by the Ethics Committee on Biomedical Research of West China Hospital of Sichuan University (Approval No. 2020-832). The data collection for this study was performed within a comprehensive, ongoing prospective cohort framework at our institution. This framework is supported by several foundational ethics approvals from the same committee, including the “colorectal cancer database application analysis” (Approval No. 2019-140), “value-based healthcare-oriented full-lifecycle cohort study of colorectal cancer” (Approval No. 2021-155), and “multicenter data cohort construction and data mining based on value-based healthcare colorectal cancer standard dataset” (Approval No. 2023-669).

All participants provided written informed consent prior to enrollment in the study.

RESULTS

Patient enrollment and baseline characteristics

From June 2018 to January 2024, 498 patients with rectal cancer who underwent sphincter-preserving surgery were enrolled in DACCA. Of these, 425 patients who met the inclusion criteria were initially included. Twenty-two patients who met the exclusion criteria (patients lost to follow-up, n = 20; death, n = 2) were excluded from the study.

Of the 403 eligible patients, 114 were assigned to the ultra-low group (≤ 3 cm), and 289 were assigned to the non-ultra-low group (> 3 cm). In the pre-matched cohort, significant baseline differences were observed between the groups. Specifically, patients in the ultra-low group were younger (P = 0.005), they had a more advanced TNM stage (P = 0.001), and they were significantly more likely to have received neoadjuvant chemoradiotherapy (P < 0.001). Other baseline characteristics, including sex, BMI, comorbidities (diabetes, hypertension), and tumor histology, were comparable between the groups (all P > 0.05, Table 1).

Table 1 Comparison of baseline characteristics in the pre-matched cohort (n = 403), n (%).

Characteristic	n	Ultra-low group (≤ 3 cm; n = 114)	Non-ultra-low group (> 3 cm; n = 289)	Test statistic	P value
Age (year)	403	57.24 ± 11.74	60.73 ± 10.87	t = -2.839	0.005
Gender
Female	168	42 (36.8)	126 (43.6)	χ2 = 1.535	0.215
Male	235	72 (63.2)	163 (56.4)
BMI (kg/m2)		23.67 ± 3.21	23.33 ± 3.02	t = 1.008	0.314
Blood type
A	130	38 (33.3)	92 (31.8)	χ2 = 0.326	0.955
AB	36	10 (8.8)	26 (9.0)
B	91	27 (23.7)	64 (22.1)
O	146	39 (34.2)	107 (37.0)
Diabetes
No	336	95 (83.3)	241 (83.4)	χ2 = 0.000	0.989
Yes	67	19 (16.7)	48 (16.6)
Hypertension
No	281	80 (70.2)	201 (69.6)	χ2 = 0.015	0.902
Yes	122	34 (29.8)	88 (30.4)
Tumor pathology
Adenocarcinoma	364	101 (88.6)	263 (91.0)	χ2 = 0.832	0.842
Mucinous adenocarcinoma	34	11 (9.6)	23 (8.0)
Neuroendocrine carcinoma	2	1 (0.9)	1 (0.3)
Signet ring cell carcinoma	3	1 (0.9)	2 (0.7)
TNM
0	28	14 (12.3)	14 (4.8)	χ2 = 18.288	0.001
I	89	35 (30.7)	54 (18.7)
II	114	22 (19.3)	92 (31.8)
III	115	26 (22.8)	89 (30.8)
IV	57	17 (14.9)	40 (13.8)
Postoperative complications
None	384	110 (96.5)	274 (94.8)	χ2 = 8.413	0.394
Anastomotic bleeding	9	2 (1.8)	7 (2.4)
Anastomotic leaks including invisible leaks	2	0 (0.0)	2 (0.7)
Stoma obstruction or intestinal obstruction	1	1 (0.9)	0 (0.0)
Incisional infection or incisional hernia	5	1 (0.9)	4 (1.4)
Other: e.g. high blood pressure, cardiac arrhythmia, deep vein thrombosis of the lower limbs	2	0 (0.0)	2 (0.6)
Neoadjuvant radiotherapy
No	287	47 (41.2)	240 (83.0)	χ2 = 69.738	< 0.001
Yes	116	67 (58.8)	49 (17.0)
Neoadjuvant chemotherapy
No	169	26 (22.8)	143 (49.5)	χ2 = 23.888	< 0.001
Yes	234	88 (77.2)	146 (50.5)

BMI: Body mass index.

To control for selection bias, 1:1 PSM was performed, yielding a matched cohort of 110 patients in each group. Post-matching, most baseline variables, including age, sex, BMI, and comorbidities, were well balanced between the groups (Table 2). However, significant differences in the rates of neoadjuvant chemoradiotherapy and TNM stage distribution persisted. This was clinically anticipated, as these factors are intrinsically linked to the tumor location. Therefore, to account for this remaining imbalance, both neoadjuvant therapy status and TNM stage were retained as key covariates in the subsequent multivariable GEE and logistic regression models to isolate the independent effect of the tumor location.

Table 2 Comparison of baseline characteristics in the matched cohort (n = 220), n (%).

Characteristic	n	Ultra-low group (≤ 3 cm; n = 110)	Non-ultra-low group (> 3 cm; n = 110)	Test statistic	P value
Age (year)	220	58.08 ± 10.978	64.47 ± 11.242	t = 4.266	0.576
Gender
Female	95	41 (37.3)	54 (49.1)	χ2 = 3.131	0.077
Male	125	69 (62.7)	56 (50.9)
BMI (kg/m2)		23.75 ± 3.14	23.23 ± 2.99	t = -1.259	0.933
Blood type
A	74	37 (33.6)	37 (33.6)	χ2 = 1.530	0.675
AB	16	10 (9.1)	6 (5.5)
B	50	26 (23.6)	24 (21.8)
O	80	37 (33.6)	43 (39.1)
Diabetes
No	176	91 (82.7)	85 (77.3)	χ2 = 1.023	0.312
Yes	44	19 (17.3)	25 (22.7)
Hypertension
No	150	76 (69.1)	74 (67.3)	χ2 = 0.084	0.772
Yes	70	34 (30.9)	36 (32.7)
Tumor pathology
Adenocarcinoma	202	97 (88.2)	105 (95.5)	χ2 = 4.583	0.205
Mucinous adenocarcinoma	15	11 (10.0)	4 (3.6)
Neuroendocrine carcinoma	1	1 (0.9)	0 (0.0)
Signet ring cell carcinoma	2	1 (0.9)	1 (0.9)
TNM
0	27	13 (11.8)	14 (12.7)	χ2 = 25.940	< 0.001
I	82	32 (29.1)	50 (45.5)
II	57	22 (20.0)	35 (31.8)
III	33	26 (23.6)	7 (6.4)
IV	21	17 (15.5)	4 (3.6)
Postoperative complications
No	210	106 (96.4)	104 (94.5)	χ2 = 0.419	0.517
Yes	10	4 (3.6)	6 (5.5)
Neoadjuvant radiotherapy
No	136	46 (41.8)	90 (81.8)	χ2 = 37.283	< 0.001
Yes	84	64 (58.2)	20 (18.2)
Neoadjuvant chemotherapy
No	102	26 (23.6)	76 (69.1)	χ2 = 45.696	< 0.001
Yes	118	84 (76.4)	34 (30.9)

BMI: Body mass index.

LARS incidence and risk at specific time points

Univariate analysis of the matched cohort revealed a significantly higher incidence of LARS in the ultra-low group than in the non-ultra-low group at all postoperative time points (Table 3). For example, the incidence at 1 month was 75.5% in the ultra-low group vs 51.8% in the non-ultra-low group (P < 0.001), and this difference persisted at 12 months (30.0% vs 14.5%, P = 0.006).

Table 3 Univariate analysis of low anterior resection syndrome occurrence in the matched cohort (n = 220), n (%).

Time point and LARS occurrence	1 month		3 months		6 months		9 months		12 months
	Yes	No	Yes	No	Yes	No	Yes	No	Yes	No
Total	140 (63.6)	80 (36.3)	101 (45.9)	119 (54.1)	88 (40.0)	132 (60.0)	62 (8.2)	158 (71.8)	49 (22.3)	171 (77.7)
Ultra-low group	83 (75.5)	27 (24.5)	64 (58.2)	46 (41.8)	56 (50.9)	54 (49.1)	39 (35.5)	71 (64.5)	33 (30.0)	77 (70.0)
Non-ultra-low group	57 (51.8)	53 (48.2)	37 (33.6)	73 (66.4)	32 (29.1)	78 (70.9)	23 (20.9)	87 (79.1)	16 (14.5)	94 (85.5)
χ2	13.279		13.344		10.909		5.749		7.588
Coefficient of contingency	0.239		0.239		0.217		0.160		0.183
P value	< 0.001		< 0.001		0.001		0.016		0.006

LARS: Low anterior resection syndrome.

To specifically address the post-PSM imbalances in neoadjuvant chemoradiotherapy and TNM stage, we then conducted separate multivariable logistic regression analyses for each of the five time points. These models confirmed that an ultra-low tumor location was a persistent and independent risk factor for LARS throughout the first year after surgery. After adjusting for all covariates, including the imbalanced factors, the ORs for LARS in the ultra-low group remained significant at every time point (e.g., 1 month, OR = 2.280, 95%CI: 1.111-4.678; 12 months, OR = 2.935, 95%CI: 1.275-6.756; all P < 0.05, Tables 4, 5, 6, 7 and 8).

Table 4 Multivariable logistic regression analysis of risk factors for low anterior resection syndrome at 1 month postoperatively.

	β	SE	Wald χ2	df	P value	Exp(B)	95%CI for Exp(B)
							Lower	Upper
Tumor location (> 3 cm)	0.824	0.367	5.055	1	0.025	2.280	1.111	4.678
Gender (female)	0.210	0.321	0.425	1	0.514	1.233	0.657	2.315
Age (< 61.28 years)	-0.483	0.331	2.127	1	0.145	0.617	0.322	1.181
Blood type (A)			1.964	3	0.580
AB	-0.582	0.417	1.952	1	0.162	0.559	0.247	1.264
B	-0.237	0.639	0.137	1	0.711	0.789	0.225	2.762
O	-0.270	0.365	0.544	1	0.461	0.764	0.373	1.563
BMI (< 23.42 kg/m2)	0.330	0.315	1.092	1	0.296	1.390	0.749	2.579
Diabetes (no)	0.232	0.401	0.334	1	0.563	1.261	0.575	2.765
Hypertension (no)	-0.224	0.339	0.436	1	0.509	0.799	0.411	1.553
Tumor pathology (adenocarcinoma)			1.270	3	0.736
Neuroendocrine carcinoma	19.502	40192.969	0.000	1	1.000	294868049.288	0.000
Signet ring cell carcinoma	19.821	28180.122	0.000	1	0.999	405570330.801	0.000
Mucinous adenocarcinoma	-0.820	0.728	1.270	1	0.260	0.440	0.106	1.834
TNM (0)			1.333	4	0.856
I	-0.127	0.522	0.059	1	0.808	0.881	0.317	2.449
II	-0.150	0.548	0.075	1	0.785	0.861	0.294	2.521
III	-0.356	0.633	0.315	1	0.574	0.701	0.203	2.424
IV	0.442	0.761	0.337	1	0.562	1.555	0.350	6.912
Postoperative complications (no)	-0.574	0.699	0.674	1	0.412	0.563	0.143	2.216
Neoadjuvant chemotherapy (no)	0.197	0.442	0.199	1	0.656	1.217	0.512	2.893
Neoadjuvant radiotherapy (no)	-0.177	0.460	0.149	1	0.700	0.838	0.340	2.062
Constant	1.273	0.928	1.883	1	0.170	3.572

Age and body mass index were converted into binary variables using the mean as the cutoff. Reference categories are shown in parentheses. BMI: Body mass index.

Table 5 Multivariable logistic regression analysis of risk factors for low anterior resection syndrome at 3 months postoperatively.

	β	SE	Wald χ2	df	P value	Exp(B)	95%CI for Exp(B)
							Lower	Upper
Tumor location (> 3 cm)	1.166	0.362	10.398	1	0.001	3.21	1.58	6.52
Gender (female)	0.425	0.317	1.799	1	0.18	1.529	0.822	2.843
Age (< 61.28 years)	-0.217	0.323	0.451	1	0.502	0.805	0.427	1.516
Blood type (A)			1.618	3	0.655
AB	-0.251	0.402	0.389	1	0.533	0.778	0.354	1.712
B	-0.476	0.614	0.601	1	0.438	0.621	0.187	2.069
O	-0.424	0.356	1.416	1	0.234	0.654	0.326	1.316
BMI (< 23.42 kg/m2)	0.078	0.309	0.064	1	0.801	1.081	0.59	1.98
Diabetes (no)	0.214	0.386	0.308	1	0.579	1.239	0.582	2.637
Hypertension (no)	-0.309	0.334	0.858	1	0.354	0.734	0.382	1.412
Tumor pathology (adenocarcinoma)			0.357	3	0.949
Neuroendocrine carcinoma	-21.921	40192.97	0	1	1	0	0
Signet ring cell carcinoma	0.172	1.63	0.011	1	0.916	1.188	0.049	29.01
Mucinous adenocarcinoma	-0.324	0.611	0.282	1	0.595	0.723	0.218	2.393
TNM (0)			8.403	4	0.078
I	0.003	0.507	0	1	0.995	1.003	0.371	2.71
II	-0.522	0.534	0.954	1	0.329	0.594	0.208	1.691
III	-1.18	0.614	3.698	1	0.054	0.307	0.092	1.023
IV	-1.258	0.673	3.495	1	0.062	0.284	0.076	1.063
Postoperative complications (no)	-0.427	0.71	0.361	1	0.548	0.653	0.162	2.625
Neoadjuvant chemotherapy (no)	0.572	0.432	1.753	1	0.185	1.772	0.76	4.131
Neoadjuvant radiotherapy (no)	-0.407	0.435	0.873	1	0.35	0.666	0.284	1.562
Constant	-0.027	0.825	0.001	1	0.974	0.973

Age and body mass index were converted into binary variables using the mean as the cutoff. Reference categories are shown in parentheses. BMI: Body mass index.

Table 6 Multivariable logistic regression analysis of risk factors for low anterior resection syndrome at 6 months postoperatively.

	β	SE	Wald χ2	df	P value	Exp(B)	95%CI for Exp(B)
							Lower	Upper
Tumor location (> 3 cm)	0.953	0.363	6.896	1	0.009	2.594	1.273	5.284
Gender (female)	-0.161	0.319	0.255	1	0.614	0.851	0.455	1.592
Age (< 61.28 years)	-0.097	0.323	0.09	1	0.764	0.908	0.482	1.71
Blood type (A)			2.487	3	0.478
AB	-0.118	0.409	0.083	1	0.773	0.889	0.399	1.981
B	0.891	0.623	2.048	1	0.152	2.438	0.719	8.264
O	0.114	0.357	0.102	1	0.749	1.121	0.557	2.257
BMI (< 23.42 kg/m2)	-0.194	0.313	0.384	1	0.535	0.824	0.447	1.52
Diabetes (no)	0.134	0.394	0.115	1	0.734	1.143	0.528	2.472
Hypertension (no)	-0.767	0.345	4.952	1	0.026	0.464	0.236	0.913
Tumor pathology (adenocarcinoma)			0.002	3	1
Neuroendocrine carcinoma	-21.827	40192.97	0	1	1	0	0
Signet ring cell carcinoma	-19.705	26717.346	0	1	0.999	0	0
Mucinous adenocarcinoma	-0.027	0.604	0.002	1	0.965	0.974	0.298	3.183
TNM (0)			2.583	4	0.63
I	0.042	0.5	0.007	1	0.933	1.043	0.391	2.779
II	-0.224	0.526	0.182	1	0.67	0.799	0.285	2.242
III	-0.285	0.598	0.228	1	0.633	0.752	0.233	2.426
IV	-0.88	0.683	1.66	1	0.198	0.415	0.109	1.582
Postoperative complications (no)	-0.351	0.759	0.213	1	0.644	0.704	0.159	3.118
Neoadjuvant chemotherapy (no)	-0.013	0.444	0.001	1	0.977	0.987	0.414	2.356
Neoadjuvant radiotherapy (no)	0.427	0.433	0.974	1	0.324	1.533	0.656	3.582
Constant	-0.485	0.822	0.348	1	0.555	0.616

Age and body mass index were converted into binary variables using the mean as the cutoff. Reference categories are shown in parentheses. BMI: Body mass index.

Table 7 Multivariable logistic regression analysis of risk factors for low anterior resection syndrome at 9 months postoperatively.

	β	SE	Wald χ2	df	P value	Exp(B)	95%CI for Exp(B)
							Lower	Upper
Tumor location (> 3 cm)	0.859	0.386	4.958	1	0.026	2.362	1.108	5.032
Gender (female)	-0.027	0.337	0.006	1	0.937	0.974	0.503	1.883
Age (< 61.28 years)	0.083	0.348	0.057	1	0.811	1.087	0.55	2.148
Blood type (A)			2.662	3	0.447
AB	-0.629	0.443	2.018	1	0.155	0.533	0.224	1.27
B	0.155	0.614	0.064	1	0.8	1.168	0.351	3.89
O	-0.345	0.374	0.856	1	0.355	0.708	0.34	1.472
BMI (< 23.42 kg/m2)	-0.096	0.331	0.083	1	0.773	0.909	0.475	1.74
Diabetes (no)	0.566	0.405	1.959	1	0.162	1.762	0.797	3.893
Hypertension (no)	-0.383	0.361	1.122	1	0.289	0.682	0.336	1.384
Tumor pathology (adenocarcinoma)			0.003	3	1
Neuroendocrine carcinoma	-20.742	40192.97	0	1	1	0	0
Signet ring cell carcinoma	-19.65	27972.494	0	1	0.999	0	0
Mucinous adenocarcinoma	0.038	0.665	0.003	1	0.955	1.038	0.282	3.821
TNM (0)			1.483	4	0.83
I	-0.064	0.512	0.015	1	0.901	0.938	0.344	2.561
II	-0.486	0.555	0.766	1	0.382	0.615	0.207	1.827
III	-0.427	0.621	0.474	1	0.491	0.652	0.193	2.202
IV	-0.428	0.699	0.375	1	0.54	0.652	0.166	2.563
Postoperative complications (no)	0.107	0.75	0.02	1	0.886	1.113	0.256	4.845
Neoadjuvant chemotherapy (no)	-0.303	0.479	0.401	1	0.526	0.738	0.289	1.887
Neoadjuvant radiotherapy (no)	0.407	0.466	0.763	1	0.382	1.503	0.603	3.747
Constant	-0.911	0.886	1.056	1	0.304	0.402

Age and body mass index were converted into binary variables using the mean as the cutoff. Reference categories are shown in parentheses. BMI: Body mass index.

Table 8 Multivariable logistic regression analysis of risk factors for low anterior resection syndrome at 12 months postoperatively.

	β	SE	Wald χ2	df	P value	Exp(B)	95%CI for Exp(B)
							Lower	Upper
Tumor location (> 3 cm)	1.077	0.425	6.41	1	0.011	2.935	1.275	6.756
Gender (female)	-0.038	0.368	0.011	1	0.918	0.963	0.468	1.981
Age (< 61.28 years)	0.004	0.38	0	1	0.992	1.004	0.476	2.115
Blood type (A)			5.134	3	0.162
AB	-0.934	0.52	3.222	1	0.073	0.393	0.142	1.09
B	0.616	0.638	0.933	1	0.334	1.851	0.53	6.459
O	-0.173	0.405	0.182	1	0.669	0.841	0.38	1.861
BMI (< 23.42 kg/m2)	-0.27	0.365	0.548	1	0.459	0.763	0.373	1.56
Diabetes (no)	0.466	0.449	1.081	1	0.299	1.594	0.662	3.841
Hypertension (no)	-0.438	0.396	1.226	1	0.268	0.645	0.297	1.401
Tumor pathology (adenocarcinoma)			1.239	3	0.744
Neuroendocrine carcinoma	-21.555	40192.97	0	1	1	0	0
Signet ring cell carcinoma	-19.393	28190.679	0	1	0.999	0	0
Mucinous adenocarcinoma	-0.73	0.655	1.239	1	0.266	0.482	0.133	1.742
TNM (0)			2.71	4	0.608
I	0.611	0.614	0.992	1	0.319	1.843	0.553	6.135
II	0.573	0.642	0.797	1	0.372	1.773	0.504	6.239
III	-0.157	0.739	0.045	1	0.832	0.855	0.201	3.635
IV	0.105	0.801	0.017	1	0.895	1.111	0.231	5.344
Postoperative complications (no)	-0.098	0.877	0.012	1	0.911	0.907	0.162	5.061
Neoadjuvant chemotherapy (no)	-0.35	0.518	0.458	1	0.499	0.705	0.255	1.943
Neoadjuvant radiotherapy (no)	0.494	0.503	0.964	1	0.326	1.638	0.611	4.39
Constant	-1.162	0.951	1.493	1	0.222	0.313

Age and body mass index were converted into binary variables using the mean as the cutoff. Reference categories are shown in parentheses. BMI: Body mass index.

Longitudinal analysis of the LARS recovery trend

To evaluate the overall effect of the tumor location on LARS risk and its longitudinal trend over the first postoperative year, a GEE model was constructed, incorporating data from all five follow-up time points (Tables 9 and 10).

Table 9 Summary of generalized estimating equation analysis for overall group, time, and interaction effects on low anterior resection syndrome.

	Wald χ2	P value
Group effect	19.239	< 0.001
Time effect	86.309	< 0.001
Interaction effect	1.063	0.900

Table 10 Multivariable generalized estimating equation parameter estimates for predictors of low anterior resection syndrome occurrence over 12 months.

	β	95%CI		Wald χ2	P value	Exp(B)	95%CI for Exp(B)
		Lower	Upper				Lower	Upper
	0.073	-0.301	0.447	0.145	0.703	1.075	0.74	1.563
Tumor distance from dentate line > 3 cm	0					1
Tumor distance from dentate line ≤ 3 cm	1.05	0.477	1.623	12.901	< 0.001	2.858	1.611	5.070
Post-operative time (month)
1	0					1
3	-0.752	-1.154	-0.35	13.456	< 0.001	0.471	0.315	0.704
6	-0.964	-1.417	-0.511	17.372	< 0.001	0.381	0.242	0.600
9	-1.403	-1.94	-0.866	26.237	< 0.001	0.246	0.144	0.421
12	-1.843	-2.465	-1.222	33.832	< 0.001	0.158	0.085	0.295

This generalized estimating equation model was adjusted for the following covariates: Age, sex, body mass index, neoadjuvant chemoradiotherapy, and TNM stage. The reference group for the “group” effect is the non-ultra-low group (> 3 cm). The reference time point for the “time” effect is 1 month postoperatively.

After adjusting for covariates, the model confirmed a significant group effect (Wald χ² = 19.239, P < 0.001), as patients in the ultra-low group had 2.858-fold higher overall odds of experiencing LARS during the first year than those in the non-ultra-low group. A significant time effect was also observed (Wald χ² = 86.309, P < 0.001), indicating that the risk of LARS substantially decreased over time for both groups. Specifically, the odds of LARS at 12 months was 0.158-fold that at the 1-month baseline.

Critically, the group-by-time interaction effect was not statistically significant (Wald χ² = 1.063, P = 0.900). This indicates that although the ultra-low group started from a much worse baseline, the rate of functional recovery itself closely paralleled that of the non-ultra-low group.

To visually represent these dynamic changes, we generated a plot of the estimated marginal means, which illustrated the predicted probability of LARS for each group across all five postoperative time points (Figure 2). The ultra-low group started with a markedly higher probability of LARS than the non-ultra-low group. Both groups exhibited a steep decline in LARS probability during the first 6 months, followed by a more gradual decrease at later points. Crucially, the two trend lines progressed in a near-parallel fashion, reinforcing the GEE model’s findings.

Figure 2 Predicted probabilities of low anterior resection syndrome in the ultra-low and the non-ultra-low rectal cancer group based on the generalized estimating equation model. LARS: Low anterior resection syndrome.

DISCUSSION

The increasing demand for sphincter-preserving surgery in rectal cancer, coupled with the inevitable impact of such procedures on anal function, has positioned research on LARS as critical for improving postoperative QoL. Our study approached this challenge by focusing on tumor location as a primary entry point. In this prospective investigation, we characterized LARS at multiple continuous observation points within the first postoperative year. This design allowed us to assess the effects of varying tumor locations on LARS at distinct time points and understand the potential differences in these effects between different time points. Moreover, as LARS is a longitudinally assessed outcome, patients inherently exhibit a recovery trend. The central question we sought to answer was whether this longitudinal trajectory is also influenced by the tumor location.

This prospective, propensity score-matched cohort study yielded two principal findings. First, we confirmed that an ultra-low tumor location is a robust and independent risk factor for LARS throughout the first postoperative year. Second, and arguably our most critical finding, we demonstrated that although the ultra-low group had a significantly higher baseline probability of LARS, the functional recovery trend of this group closely paralleled that of the non-ultra-low group.

Ultra-low tumor location as a persistent risk factor for LARS throughout the first postoperative year

We chose the tumor location as a primary analytical focus because it represents a direct proxy for the resulting anastomotic height, which is a critical determinant of postoperative anal function. The mechanisms linking low anastomoses to LARS are well established. Specifically, surgery for ultra-low tumors (in our study, defined as ≤ 3 cm from the dentate line) necessitates resection near the anal canal. This procedure inherently reduces the rectal reservoir volume[24], risks damage to essential autonomic nerves (e.g., superior hypogastric plexus), thereby impairing motility[25,26], and potentially ablates sensory receptors in the anorectal transition zone, leading to neurocontrol dysfunction and urgency[27,28]. Therefore, a strong association between an ultra-low tumor location and a higher risk of LARS is mechanistically plausible and expected.

Our findings robustly validated this hypothesis. The results of our separate multivariable logistic regression models confirmed that an ultra-low tumor location was a persistent and independent risk factor for LARS throughout the entire first postoperative year (all P < 0.05). This finding is broadly consistent with existing research. Previous cross-sectional studies and meta-analyses similarly identified low tumor height (variously defined as < 6 cm, < 5 cm, or < 4 cm from the anal verge or in relation to remnant rectal volume) as a primary risk factor for major LARS[29-31].

However, our study significantly expands this existing knowledge. By employing a longitudinal design with five distinct follow-up points, we demonstrated that this elevated risk is a persistent state that continues throughout the first year of recovery. Furthermore, our multiple-timepoint analysis revealed a nuanced picture: Although the risk is persistent, its magnitude varies. As indicated by the differing ORs at the observation points (Tables 4, 5, 6, 7 and 8), the impact of the tumor location on the probability of LARS is not static, suggesting dynamic interplay between the initial surgical trauma and the ongoing recovery process.

LARS risk progressively decreases postoperatively, regardless of the initial tumor location

The dynamic trajectory of LARS is of paramount interest, as it reflects the patient’s real-world QoL improvement more accurately than a single static assessment. Our study captured this as a comprehensive recovery state at five time points, which inherently includes the effects of any unmeasured clinical interventions or lifestyle modifications. We utilized GEE analysis to accurately model these longitudinal data and account for correlations within the same patient.

This analysis yielded two innovative and clinically significant findings. First, the GEE model confirmed a highly significant “time effect” (P < 0.001), demonstrating a sustained recovery trend for all patients. The overall odds of LARS at 12 months decreased to 0.158-fold that at the 1-month baseline. This universal improvement, likely driven by mechanisms such as intestinal adaptation and neural compensation[32-37], is a crucial finding for clinicians and patients. It provides robust evidence that even patients with a high-risk ultra-low tumor location can expect continuous functional improvement, which is vital for bolstering patient confidence and justifying active rehabilitation efforts.

Second, and arguably the most valuable finding of this study, is the comparison of these recovery trends. The GEE analysis demonstrated a non-significant group-by-time interaction (P = 0.900). This indicates that the rate of functional improvement was statistically similar between the groups. This finding implies that although the initial degree of surgical trauma (dictated by the tumor location) sets the severe baseline risk, the subsequent recovery mechanisms themselves operate effectively regardless of the starting point[38-40]. For patients in the ultra-low group, who have a severe initial symptom burden, their similar recovery rate as their counterparts in the non-ultra-low group might translate to a more pronounced subjective sense of relief and improvement.

These parallel recovery trajectories imply that the fundamental processes of tissue repair and neural plasticity might be governed by conserved systemic or local mechanisms. Recent tumor biology research has underscored that the tumor microenvironment plays a pivotal role in tissue homeostasis, inflammation, and repair. For instance, extracellular vesicle-mediated exchange of miRNAs between tumor cells and cancer-associated fibroblasts has been found to regulate pathways involved in fibrosis, angiogenesis, and immune modulation in other cancers[41]. Although our study did not measure such effectors, it is plausible that similar extracellular vesicle-mediated or paracrine signaling within the pelvic tissue milieu post-resection might influence the pace of neural adaptation and tissue remodeling. The parallel recovery observed in this study suggests that such pro-repair mechanisms, once initiated, can operate with comparable efficiency regardless of the initial extent of resection, offering a novel biological lens through which to view functional rehabilitation.

However, this population-level parallelism likely overlays significant individual variation in the pace and completeness of functional recovery. An individual’s genetic constitution is a key modulator of such variation, influencing tolerance to therapeutic trauma and regenerative capacity. This concept is exemplified by hereditary syndromes such as Lynch syndrome, in which specific germline mutations both predispose to colorectal cancer and appear to influence the incidence and severity of side effects from antineoplastic treatments[42]. Although our study was not designed to assess genetic markers, this framework suggests that within both the ultra-low and non-ultra-low groups, unmeasured genetic factors could partly explain why some patients experienced rapid resolution of LARS whereas others exhibited protracted symptoms despite similar anatomical starting points. Thus, although tumor location robustly sets the group-level baseline risk and recovery curve, future models aiming for true personalization must integrate individual-level factors, including genetic susceptibility, to predict and optimally manage post-treatment functional outcomes.

Nevertheless, although these parallel trajectories do not imply equivalent outcomes, the findings carry important clinical implications. Patients in the ultra-low group consistently exhibited a higher absolute risk at every time point because of their significantly worse baseline[43]. This persistent gap (e.g., 30.0% vs 14.5% at 12 months) necessitates more early and intensive intervention, such as systematic gastrointestinal rehabilitation protocols[41,44], to compensate for their elevated baseline risk and help close this functional divide.

Strengths and limitations

This study had several notable strengths. Its prospective cohort design, with a large sample size and a 12-month follow-up period involving five distinct observation points, provided a robust dataset. Methodologically, the use of a GEE model to analyze these multiple-timepoint data is a significant advantage, making this one of the few studies to rigorously model the dynamic trajectory of LARS and investigate the impact of the tumor location on this trend. Furthermore, the application of 1:1 PSM enhanced the comparability between the groups.

However, certain limitations must be acknowledged. First, despite PSM, significant baseline imbalances persisted regarding neoadjuvant chemoradiotherapy and TNM stage. We deliberately addressed this limitation by including these factors as covariates in all subsequent multivariable analyses (both the time point logistic regressions and the final GEE model). These analyses confirmed that neoadjuvant chemoradiotherapy and TNM stage were not independent predictors of LARS in our cohort, whereas tumor location remained the sole, significant independent factor (e.g., OR = 2.280 at 1 month). We are therefore confident that this residual imbalance did not compromise our primary findings. Second, similar to most studies in this field, we lacked preoperative gastrointestinal function data[45-48]. It is possible that some LARS symptoms represent pre-existing functional disorders rather than a direct consequence of surgery. Future studies incorporating preoperative assessments are needed to clarify the true incidence of LARS. Third, our primary outcome relied on the LARS score. Although validated and widely used, this tool is inherently subjective[49,50]. Studies have suggested it could have limited specificity and it could potentially over- or underestimate the true impact on QoL[51,52]. This subjectivity, common to all LARS research, might have contributed to inconsistencies in findings across studies. Fourth, surgeon experience was not recorded, and this represents a potential unmeasured confounder[53,54]. However, all surgeries were performed by senior surgeons in a specialized center using standardized techniques, which likely reduced variability. Future studies should incorporate surgeon-related variables to further clarify their role.

CONCLUSION

This study confirmed that an ultra-low rectal cancer location is a persistent, independent risk factor for LARS throughout the first postoperative year. Although all patients exhibited a universal trend of functional recovery, the longitudinal trend for the ultra-low and non-ultra-low groups closely paralleled. Consequently, the ultra-low group, beginning with a significantly worse baseline, maintained a persistently higher absolute risk of gastrointestinal dysfunction. These findings establish the tumor location as a crucial prognostic factor in rectal cancer. This knowledge should be incorporated into preoperative patient counseling to manage expectations and guide the implementation of targeted postoperative rehabilitation strategies to help close the persistent functional gap in this high-risk population. With deeper insights into tumor biology, future cancer treatment clearly move toward personalized and precision strategies[55]. Future research should integrate anatomical predictors with molecular profiling to build multidimensional models for functional outcomes. This multimodal approach to risk prediction will ultimately support truly personalized postoperative care strategies, aligning functional recovery with the overarching goals of precision medicine.