Postoperative outcomes following prehabilitation vs no prehabilitation in elective colorectal surgery: A 2:1 propensity score-matched analysis

Abstract

BACKGROUND

Aging is associated with reduced physiological reserves, frailty, sarcopenia, and increases in other comorbidities. Existing studies on prehabilitation showed potential in improving postoperative outcomes but were heterogeneous. Recommendations are still weak for promoting multimodal prehabilitation before elective colorectal surgery. There is also no recommendation on the best criteria to select patients for prehabilitation.

AIM

To compare postoperative outcomes in patients with prehabilitation vs those without prior to elective colorectal surgery using propensity score matching (PSM).

METHODS

This retrospective study was conducted from July 2010 to December 2021 on patients aged ≥ 75 years, or aged ≥ 65 years and frail, and/or had Charlson comorbidity index ≥ 4 who underwent elective colorectal surgery. Perioperative care included a specialized nurse, physiotherapist, dietician, and geriatrician review. Decision for the type of prehabilitation (inpatient, outpatient, or home-based) was made after a joint discussion considering patient’s mobility, caregiver availability, and anticipated compliance to physiotherapy regime. A 2:1 PSM was performed to balance.

RESULTS

There were 208 patients (prehabilitation: 148, no prehabilitation: 60) in the unmatched cohort. There were 43.8% who were aged ≥ 80 years, 31.7% who were frail, and 43.8% who underwent laparoscopic surgery. Duration of prehabilitation ranged from 2-4 weeks. The overall incidence of major morbidity, 30-day mortality, and 1-year mortality were 12.0%, 1.9%, and 6.7%, respectively. The PSM group had 144 patients. Prehabilitation was not an independent predictor of major morbidity [odds ratio (OR) = 0.84, 95% confidence interval (CI): 0.30-2.33] and 1-year mortality (OR = 1.26, 95%CI: 0.30-5.28). Subgroup analysis of patients who were not frail (prehabilitation n = 79, no prehabilitation n = 40) similarly showed that prehabilitation was not an independent predictor of major morbidity (OR = 0.75, 95%CI: 0.26-2.14) and 1-year mortality (OR = 1.04, 95%CI: 0.24-4.55).

CONCLUSION

Older patients who were not frail did not benefit from prehabilitation. Selection criteria for prehabilitation may be modified to target patients who will better benefit from it.

Key Words: Colorectal surgery; Elderly; Enhanced Recovery After Surgery; Frailty; Hemicolectomy; Prehabilitation

Core Tip: Aging is associated with frailty, increased comorbidities, and worse postoperative outcomes. Some studies on prehabilitation showed its potential in improving postoperative outcomes for elective colorectal surgery, but selection criteria and prehabilitation protocols are heterogeneous. Through propensity score matching, this study showed that prehabilitation in non-frail older patients (≥ 65 years) did not improve short-term outcomes and 1-year mortality. Open surgery was an independent predictor of 1-year mortality. Consideration should be made to modify and stratify the selection criteria for prehabilitation based on frailty scores and comorbidity burden to identify the group of patients who will best benefit given finite resources.

INTRODUCTION

Colorectal cancer is the third most common cancer and ranks second in terms of cancer-related mortality[1]. With advances in healthcare innovation and increased availability of healthcare, countries face a shift in demographics in which the proportion of the world’s population older than 60 years is expected to nearly double from 12% to 22% between 2015 to 2050[2]. Aging is associated with reduced physiological reserves, lean body mass, sarcopenia, and increased comorbidities[3,4]. Ninety-day mortality following colorectal surgery for colorectal cancer was reported to be higher for patients aged 50 years or more compared with those less than 50 years old (6.2% vs 1.5%)[5].

Nevertheless, while old age used to be a relative contraindication for surgery, the incorporation of Enhanced Recovery After Surgery (ERAS) protocols (e.g., choice of minimally invasive surgery, avoidance of prolonged fasting, and early postoperative mobilization and oral nutrition) have improved postoperative outcomes by minimizing surgical stress response and maintaining homeostasis[6,7]. The concept of prehabilitation was also introduced to optimize patients physically and emotionally to withstand the high surgical stress[8]. Several studies have assessed the use of various prehabilitation regimes prior to elective abdominal surgery with promising results[9].

The 2022 guidelines by the American Society of Colon and Rectal Surgeons (ASCRS) and the Society of American Gastrointestinal and Endoscopic Surgeons (SAGES) recommended the consideration of multimodal prehabilitation prior to elective colorectal surgery in patients with multiple comorbidities or significant deconditioning[10]. However, the strength of recommendation was weak (level 2B evidence), and this was attributed to the lack of power in several studies with conflicting results. Currently, in our institution patients who are ≥ 65 years with frailty or with significant comorbidities are offered exercise-based prehabilitation as part of our institution’s geriatric surgical service (GSS). However, given a finite supply of healthcare resources, it is prudent to identify the group of patients who will benefit most from any intervention.

In line with the recommendations by the ASCRS and SAGES, this study aimed to review the postoperative outcomes in patients who underwent prehabilitation to identify if there is a need to modify our selection criteria for prehabilitation to ensure that limited resources are allocated to those who may benefit more. We used propensity score matching (PSM) to balance out confounding factors.

MATERIALS AND METHODS

This was a single-center retrospective study in an urban tertiary government hospital comparing the short-term outcomes of older patients who underwent prehabilitation vs no prehabilitation (as part of our GSS) for elective colonic or rectal resection for colon or rectal cancer respectively from July 2010 to December 2021. The inclusion criteria for prehabilitation were: (1) ≥ 65 years; and (2) Frail (defined as Fried’s frailty phenotype score (FFP) ≥ 3)[11] and/or had Charlson comorbidity index (CCI) ≥ 4[12]. Patients who underwent emergency surgery, surgery with palliative intent, non-colorectal surgery, or non-bowel resection surgery were excluded from this study.

Patients in the no prehabilitation group were patients who did not fit the inclusion criteria as above, had a clinical condition that warranted more urgent surgery precluding prehabilitation (e.g., impending obstruction), or refused prehabilitation. All patients were managed by our GSS. Options of both laparoscopic and open surgery were offered to patients with laparoscopy being the main option given its associated benefits; however, some considerations for open surgery instead of laparoscopy included previous abdominal surgery (that may increase difficulty of laparoscopy), underlying comorbidities that preclude the longer duration of laparoscopic surgery/ability to tolerate pneumoperitoneum, or large tumor size with anticipated difficulty of laparoscopic surgery. This study was approved by the Institutional Review Board of National Healthcare Group Domain Specific (Approval No. 2021/00967). A detailed timeline tracing key milestones from the inception of our study is appended in Supplementary Figure 1.

Preoperative management

Patients were recruited for our GSS if they were ≥ 75 years old, ≥ 65 years old with frailty, CCI ≥ 4, and/or required at least moderate assistance in activities of daily living. The GSS is a perioperative transdisciplinary service including a team of clinicians (surgeon, geriatrician, anesthetist, and cardiologist) and allied health professionals (specialized nurse clinician, physiotherapist, and dietician) led by the colorectal surgery team that was implemented in 2007[6]. Additionally, all patients were also managed according to the standard ERAS protocol[13].

Prehabilitation is generally defined as the use of physical and psychological assessments to identify baseline function and the provision of targeted interventions (including physical, nutritional, and psychological) to reduce the incidence and severity of current and future impairments[9]; however, for the purpose of this study, prehabilitation will be defined as the use of only physical interventions (i.e. exercise-based prehabilitation). Available types of prehabilitation included inpatient prehabilitation at a community hospital, day prehabilitation in an outpatient clinic, or home-based prehabilitation. Decision for the type of prehabilitation was made through a joint discussion between the GSS team, the patient and their family members with the following considerations: Patient mobility; caregiver availability; anticipated compliance to physiotherapy regime; and the need for tighter preoperative nutritional optimization. Details of the various types of prehabilitation are described in Supplementary Table 1. Prehabilitation sessions were modified and tailored using objective markers such as the modified Barthel index (MBI), 10 m walk test, and chair-stand repetitions based on the discretion of the physiotherapist. The average duration for prehabilitation ranged between 2-4 weeks.

Study variables and outcomes

The study variables include patient demographics [gender, age, CCI, American Society of Anesthesiologists (ASA) score, MBI, frailty, preoperative ambulatory status], disease characteristics (tumor stage and site), intraoperative variables (indications for surgery, surgical access, primary anastomosis, need for stoma creation, length of operation, estimated blood loss, need for transfusion and inotropes intraoperatively), and postoperative outcomes (admission to high dependency unit or intensive care unit, major morbidity, length of hospitalization stay, discharge disposition, 30-day mortality, and 30-day readmission).

While FFP defined non-frail as FFP of 0 and pre-frail as FFP of 1-2, for the purpose of this study, we defined non-frail as FFP ≤ 2; this definition has also been used in certain studies investigating the impact of frailty on postoperative outcomes[14]. Our primary outcomes were major morbidity (defined as Clavien-Dindo ≥ grade 3A), 30-day mortality, and 1-year mortality. Our secondary outcomes were length of stay (LOS) and > 10% drop in MBI at 6 weeks postoperatively. Postoperative complication was defined as the presence of any surgical complication(s) (anastomotic leak, intra-abdominal collection, ileus) or medical complication(s) (pleural effusion, pneumonia, urinary tract infection, deep vein thrombosis). However, the specific type of postoperative complication was not collected or reported in this study.

Statistical analysis

All categorical variables were described as n (%) and were analyzed using the χ² test. All continuous variables were expressed as mean ± SD or median [interquartile range (IQR)]. To account for differences in baseline demographics between groups, PSM was performed. A nearest neighbor matching algorithm with a 2:1 ratio was applied without replacement. The propensity score was estimated using logistic regression, and PSM was performed using a caliper width of 0.2 of the standard deviation of the logit of the propensity score[15]. In the event of ties, control subjects were randomly selected. Seven variables were used in the adjustment: Age; sex; MBI; frailty; ASA ≥ 3; CCI ≥ 4; and surgical access[16-18]. For variables with missing data (n < 10) (CCI, Portsmouth-Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity predicted mortality and morbidity), mean imputation was performed.

Standardized mean difference (SMD) and Hansen and Bowers test were used to assess covariable and global imbalance, respectively[19]. Adequate covariate balance was defined as SMD < 0.100. Univariate logistic regression was performed to assess the impact of prehabilitation on postoperative outcomes in both the unmatched and matched cohorts. For covariates that did not reach adequate balance despite PSM, multivariate logistic regression was performed to allow for double adjustment in the matched group[20]. Subgroup analysis would be ideal to compare postoperative outcomes in frail patients who received prehabilitation vs no prehabilitation; however, in view of the small sample size of patients who were frail in our study, this was not performed due to the lack of statistical power.

The initial PSM was performed using R software (R-3.3.3); subsequent statistical analyses were performed using SPSS version 25 (SPSS, SPSS Inc., Chicago, IL, United States). A P value of < 0.05 was used to define statistical significance.

Sample size calculation was also performed for a parallel 2:1 group allocation with a power of 0.90 and alpha of 0.05. A sample size of 82 was required to reduce postoperative any morbidity from 40% to 20%. A sample size of 359 was required to reduce postoperative any morbidity from 40% to 30%. We chose a value of 40% for any morbidity as incidence of any morbidity in older patients undergoing elective colorectal surgery ranged from 26%-57.1% in existing literature[21-25].

RESULTS

A total of 208 patients were included in this study. The overall median age was 79.0 years (IQR 76.0-82.0). There were 103 males (49.5%) and 131 patients (63.0%) with ASA ≥ 3. The majority of patients (n = 177, 85.1%) underwent colonic resection. Prior to PSM there were significantly more patients with CCI ≥ 4 (60.1%, n = 89 vs 45.0%, n = 27; P = 0.046) and diagnosed with frailty (38.5%, n = 57 vs 15.0%, n = 9; P < 0.001) who received prehabilitation compared to those who did not. More patients received prehabilitation (near statistical significance P = 0.076) if they underwent open (76.1%, n = 89) compared with laparoscopic surgery (64.8%, n = 59). Sex, tumor stage, and tumor site were comparable between the two groups. No statistically significant difference (P = 0.479) was noted in prehabilitation rates between an open (64.0%, n = 48) and laparoscopic (69.6%, n = 48) approach following PSM.

PSM resulted in 144 patients (prehabilitation n = 96, no prehabilitation n = 48). There were seven covariates with SMD > 0.100 prior to PSM; after PSM, only two covariates (ASA ≥ 3 and surgical access) had SMD > 0.100 (Figure 1). There was improvement of balance between groups after PSM using the Hansen and Bowers test (after PSM: χ² = 1.25, P = 0.99; before PSM: χ² = 17.9, P = 0.0125). Clinical demographics were comparable between the prehabilitation and no prehabilitation group following PSM (Table 1).

Figure 1 Standardized mean difference in the covariates used before (blue) and after (red) propensity score matching. The black line indicates a standardized mean difference of 0.100. ASA: American Society of Anesthesiologists; CCI: Charlson comorbidity index; SMD: Standardized mean difference.

Table 1 Clinical demographics of patients who underwent elective major colorectal surgery in those with and without prehabilitation.

	Overall group, n = 208					PSM group, n = 144
	Total	Prehabilitation (n = 148)	No prehabilitation (n = 60)	P value	SMD	Total	Prehabilitation (n = 96)	No prehabilitation (n = 48)	P value	SMD
Age, years1	79.0 (76.0-82.0)	79.0 (76.0-82.0)	79.0 (75.3-81.0)	0.371	0.167	78.0 (75.0-82.0)	78.0 (75.0-82.0)	79.0 (76.0-82.0)	0.603	0.076
Age ≥ 80 years	91 (43.8)	66 (44.6)	25 (41.7)	0.700	N/A	57 (39.6)	36 (37.5)	21 (43.8)	0.470	N/A
Sex, male1	103 (49.5)	70 (47.3)	33 (55.0)	0.314	0.154	73 (50.7)	49 (51.0)	24 (50.0)	0.906	0.021
ASA score	3.0 (2.0-3.0)	3.0 (2.0-3.0)	3.0 (2.0-3.0)	0.091	N/A	3.0 (2.0-3.0)	3.0 (2.0-3.0)	3.0 (2.0-3.0)	0.410	N/A
≥ 31	131 (63.0)	98 (66.2)	33 (55.0)	0.129	0.230	80 (55.6)	55 (57.3)	25 (52.1)	0.553	0.104
CCI	4.0 (2.3-6.0)	4.0 (3.0-6.0)	3.0 (2.0-5.0)	0.004a	N/A	3.0 (2.0-6.0)	3.0 (2.0-6.0)	3.0 (2.0-5.0)	0.153	N/A
≥ 41	116 (55.8)	89 (60.1)	27 (45.0)	0.046a	0.305	68 (47.2)	46 (47.9)	22 (45.8)	0.813	0.042
Predicted mortality					N/A					N/A
P-POSSUM, %	6.9 (4.0-10.6)	6.8 (4.2-10.7)	7.0 (3.7-10.5)	0.465	N/A	6.2 (3.4-10.3)	6.2 (3.4-10.1)	7.0 (3.5-10.8)	0.879	N/A
P-POSSUM > 5%1	140 (67.3)	102 (68.9)	38 (63.3)	0.437	N/A	93 (64.6)	62 (64.6)	31 (64.6)	1.000	N/A
Predicted morbidity, %	41.8 (30.4-53.0)	42.6 (31.0-54.5)	39.8 (29.3-51.0)	0.199	N/A	39.0 (29.3-50.9)	38.3 (28.9-50.9)	39.8 (29.3-51.0)	0.909	N/A
Modified Barthel index1	100 (98-100)	100 (98-100)	100 (100-100)	0.009a	0.174	100 (100-100)	100 (100-100)	100 (100-100)	0.572	0.019
Premorbid status				0.014a	N/A				0.354	N/A
Community ambulant	183 (88.0)	125 (84.5)	58 (96.7)		N/A	134 (93.1)	88 (91.7)	46 (95.8)		N/A
Home bound	25 (12.0)	23 (15.5)	2 (3.3)		N/A	10 (6.9)	8 (8.3)	2 (4.2)		N/A
Bed bound	0	0	0		N/A	0	0	0		N/A
Frail, yes1	66 (31.7)	57 (38.5)	9 (15.0)	< 0.001a	0.548	25 (17.4)	17 (17.7)	8 (16.7)	0.876	0.027
Tumor stage				0.618	N/A				0.365	N/A
0	18 (8.7)	14 (9.5)	4 (6.7)		N/A	12 (8.3)	10 (10.4)	2 (4.2)		N/A
1	25 (12.0)	18 (12.2)	7 (11.7)		N/A	19 (13.2)	12 (12.5)	7 (14.6)		N/A
2	77 (37.0)	51 (34.5)	26 (43.3)		N/A	51 (35.4)	30 (31.3)	21 (43.8)		N/A
3	85 (40.9)	62 (41.9)	23 (38.3)		N/A	60 (41.7)	42 (43.8)	18 (37.5)		N/A
4	3 (1.4)	3 (2.0)	0 (0)		N/A	2 (1.4)	2 (2.1)	0 (0)		N/A
Tumor stage ≥ 3, yes	88 (42.3)	65 (43.9)	23 (38.3)	0.460	N/A	62 (43.1)	44 (45.8)	18 (37.5)	0.341	N/A
Tumor site				0.377	N/A				0.743	N/A
Colon	177 (85.1)	128 (86.5)	49 (81.7)		N/A	122 (84.7)	82 (85.4)	40 (83.3)		N/A
Rectum	31 (14.9)	20 (13.5)	11 (18.3)		N/A	22 (15.3)	14 (14.6)	8 (16.7)		N/A
Surgical access1				0.076	0.271				0.479	0.125
Open	117 (56.3)	89 (60.1)	28 (46.7)		N/A	75 (52.1)	48 (50.0)	21 (43.8)		N/A
Laparoscopic	91 (43.8)	59 (39.9)	32 (53.3)		N/A	59 (47.9)	48 (50.0)	21 (43.8)		N/A

¹Propensity score matching was performed for these variables due to potential and/or significant effects on clinical outcomes or due to significant differences in demographics between the two study groups.

^aP < 0.05.

Data are presented as median (interquartile range) or n (%). ASA: American Society of Anesthesiologists; CCI: Charlson comorbidity index; P-POSSUM: Portsmouth-Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity; SMD: Standardized mean difference; N/A: Not applicable; PSM: Propensity score matching.

Intraoperative outcomes

The overall median operating time was 221.5 (IQR 172.8-272.8) min in the unmatched group. There were no significant differences in operating time, need for intraoperative blood transfusion or inotropes, and open conversion between prehabilitation and no prehabilitation in both the unmatched and matched groups (Table 2).

Table 2 Intraoperative characteristics of patients who underwent elective major colorectal surgery.

	Overall group, N = 208				PSM group, n = 144
	Prehabilitation (n = 148)	No prehabilitation (n = 60)	OR (95%CI)1	P value1	Prehabilitation (n = 96)	No prehabilitation (n = 48)	OR (95%CI)1	P value1	OR (95%CI)2	P value2
Open conversion	9/68 (13.2)	5/37 (13.5)	0.98 (0.30, 3.16)	0.968	7/66 (12.7)	5/53 (19.2)	0.61 (0.17, 2.15)	0.442	0.60 (0.17, 2.13)3	0.4333
Primary anastomosis, yes	124 (83.8)	56 (93.3)	0.37 (0.12, 1.11)	0.068	85 (88.5)	45 (93.8)	0.52 (0.14, 1.94)	0.320	0.47 (0.12, 1.79)	0.268
Stoma creation, yes	34 (23.0)	13 (21.7)	1.08 (0.52, 2.22)	0.838	19 (19.8)	9 (18.8)	1.07 (0.44, 2.58)	0.882	1.15 (0.47, 2.83)	0.758
Operating time, minute
Mean (mean ± SD)	215.6 ± 77.3	231.0 ± 76.6	N/A	0.194	225.8 ± 64.9	223.9 ± 78.1	N/A	0.884
Median	203.5 (164.3-265.8)	228.5 (165.0-275.8)	N/A	0.139	214.0 (177.8-269.0)	225.5 (152.5-274.0)	N/A	0.760
Estimated blood loss ≥ 500 mL, yes	9 (6.1)	2 (3.3)	1.88 (0.39, 8.96)	0.422	6 (6.3)	2 (4.2)	1.53 (0.30, 7.90)	0.607	1.76 (0.33, 9.48)	0.509
Intraoperative blood transfusion	21 (14.2)	8 (13.3)	1.08 (0.45, 2.58)	0.872	12 (12.5)	7 (14.6)	0.84 (0.31, 2.28)	0.728	0.88 (0.32, 2.46)	0.812
Intraoperative inotropes use	2 (1.4)	2 (3.3)	0.40 (0.06, 2.89)	0.346	2 (2.1)	1 (2.1)	1.00 (0.09, 11.31)	1.000	1.09 (0.09, 12.78)	0.945

¹Univariate analysis.

²Multivariate analysis.

³Multivariate analysis excluded the variable (surgical access).

Data are presented as median (interquartile range) or n (%). OR: Odds ratio; CI: Confidence interval; N/A: Not applicable; SD: Standard deviation; PSM: Propensity score matching.

Postoperative outcomes

The postoperative outcomes in both the unmatched and matched groups are detailed in Table 3. In the unmatched group there were 25 (12.0%), 4 (1.9%), and 14 (6.7%) patients with major morbidity, 30-day mortality, and 1-year mortality, respectively. Univariate logistic regression showed that prehabilitation did not have a significant impact on major morbidity, LOS, 30-day readmission, 30-day mortality, and 1-year mortality as well as patients with > 10% drop in MBI at 6 weeks following surgery. These results were similarly noted in the PSM group.

Table 3 Postoperative outcomes in patients who received prehabilitation vs those who did not receive prehabilitation.

	Overall group, N = 208				PSM group, n = 144
	Prehabilitation (n = 148)	No prehabilitation (n = 60)	OR (95%CI)1	P value1	Prehabilitation (n = 96)	No prehabilitation (n = 48)	OR (95%CI)1	P value1	OR (95%CI)2	P value2
Admission to ICU/HDU	59 (39.9)	19 (31.7)	1.43 (0.76, 2.70)	0.269	35 (36.5)	16 (33.3)	1.15 (0.55, 2.38)	0.712	1.19 (0.54, 2.63)	0.665
Unplanned	8 (13.6)	5 (26.3)	0.44 (0.12, 1.56)	0.194	6 (17.1)	5 (31.3)	0.46 (0.12, 1.80)	0.256	0.55 (0.13, 2.31)	0.417
Postoperative complications
Major morbidity	17 (11.5)	8 (13.3)	0.84 (0.34, 2.07)	0.711	12 (12.5)	7 (14.6)	0.84 (0.31, 2.28)	0.728	0.84 (0.30, 2.33)	0.735
Grade 3	8 (5.4)	4 (6.7)	0.80 (0.23, 2.76)	0.724	6 (6.3)	4 (8.3)	0.73 (0.20, 2.73)	0.643	0.71 (0.19, 2.68)	0.617
Grade 4	7 (4.7)	2 (3.3)	1.44 (0.29, 7.14)	0.654	4 (4.2)	2 (4.2)	1.00 (0.18, 5.66)	1.000	1.01 (0.17, 6.06)	0.993
Grade 5	2 (1.4)	2 (3.3)	0.40 (0.06, 2.89)	0.346	2 (2.1)	1 (2.1)	1.00 (0.09, 11.31)	1.000	0.93 (0.08, 11.19)	0.953
Length of stay, days
Mean (mean ± SD)	9.9 ± 9.8	9.3 ± 7.4	N/A	0.670	9.8 ± 10.5	9.1 ± 6.6	N/A	0.616	N/A	N/A
Median	7.0 (5.0-11.0)	7.0 (5.0-11.0)	N/A	0.349	7.0 (5.0-11.0)	7.0 (4.3-10.8)	N/A	0.835	N/A	N/A
≥ 14 days	28 (18.9)	13 (21.7)	0.84 (0.40, 1.77)	0.652	18 (18.8)	10 (20.8)	0.88 (0.37, 2.08)	0.766	0.87 (0.36, 2.11)	0.754
Discharge disposition			N/A	0.359			N/A	0.793	N/A	N/A
Home	121 (82.3)	52 (89.7)	N/A	N/A	86 (90.5)	41 (87.2)	N/A	N/A	N/A	N/A
Community hospital	24 (16.3)	5 (8.6)	N/A	N/A	8 (8.4)	5 (10.6)	N/A	N/A	N/A	N/A
Nursing home	2 (1.4)	1 (1.7)	N/A	N/A	1 (1.1)	1 (2.1)	N/A	N/A	N/A	N/A
30-day readmission	16 (10.9)	11 (18.6)	0.53 (0.23, 1.23)	0.136	9 (9.5)	8 (16.7)	0.52 (0.19, 1.46)	0.209	0.53 (0.19, 1.49)	0.230
Mortality
30-day	2 (1.4)	2 (3.3)	0.40 (0.06, 2.89)	0.346	2 (2.1)	1 (2.1)	1.00 (0.09, 11.31)	1.000	0.93 (0.08, 11.19)	0.953
1-year	10 (6.8)	4 (6.7)	1.01 (0.31, 3.37)	0.981	7 (7.3)	3 (6.3)	1.18 (0.29, 4.78)	0.817	1.26 (0.30, 5.28)	0.756
MBI at 6 weeks postoperatively	100 (87.8-100)	100 (98.0-100)	N/A	0.069	100 (98.0-100)	100 (94.5-100)	N/A	0.777	N/A	N/A
> 10% drop in MBI 6 weeks after surgery, yes	18 (12.2)	5 (8.3)	1.52 (0.54, 4.31)	0.425	9 (9.4)	4 (8.3)	1.14 (0.33, 3.90)	0.837	1.18 (0.34, 4.09)	0.797

¹Univariate analysis.

²Multivariate analysis.

Data are presented as median (interquartile range) or n (%). ICU: Intensive care unit; HDU: High dependency unit; MBI: Modified Barthel index; OR: Odds ratio; CI: Confidence interval; N/A: Not applicable; PSM: Propensity score matching; SD: Standard deviation.

Multivariate analysis was also performed to include ASA ≥ 3 and surgical access as covariates since their respective SMDs were still > 0.100 following PSM. Nevertheless, prehabilitation was not an independent predictor of major morbidity, LOS, 30-day readmission, 30-day mortality, 1-year mortality, and > 10% drop in MBI at 6 weeks following surgery. However, open surgery was an independent predictor of any morbidity [odds ratio (OR) = 2.71, 95% confidence interval (CI): 1.35-5.48, P = 0.005] and 1-year mortality (OR = 9.12, 95%CI: 1.12-74.32, P = 0.039). ASA ≥ 3 was an independent predictor of any morbidity (OR = 2.02, 95%CI: 1.01-4.06, P = 0.048).

Subgroup analysis was performed in patients who were not frail in both the unmatched and matched groups (Table 4). The incidence of major morbidity (13.9%, n = 11 vs 17.5%, n = 7; P = 0.588), 30-day mortality (1.3%, n = 1 vs 2.5%, n = 1; P = 0.574), 1-year mortality (7.6%, n = 6 vs 7.5%, n = 3; P = 0.961), and LOS (9.86 ± 11.20 days vs 9.20 ± 7.04 days, P = 0.734) were comparable between prehabilitation and no prehabilitation in the matched group. Subgroup analyses were also performed in the matched cohort for patients with CCI ≥ 4 as well as those who underwent open surgery only; no significant differences in major morbidity, LOS, 30-day mortality, and 1-year mortality were observed between the prehabilitation and no prehabilitation groups.

Table 4 Postoperative outcomes in the subgroup analysis of patients who were not frail.

	Overall group, N = 142				PSM group, n = 119
	Prehabilitation (n = 91)	No prehabilitation (n = 51)	OR (95%CI)1	P value1	Prehabilitation (n = 79)	No prehabilitation (n = 40)	OR (95%CI)1	P value1	OR (95%CI)2	P value2
Admission to ICU/HDU	41 (45.1)	15 (29.4)	1.97 (0.95, 4.09)	0.067	33 (41.8)	13 (32.5)	1.49 (0.67, 3.31)	0.326	1.50 (0.63, 3.57)	0.360
Unplanned	7 (17.1)	4 (26.7)	0.57 (0.14, 2.31)	0.424	6 (18.2)	4 (30.8)	0.50 (0.12, 2.18)	0.351	0.57 (0.12, 2.63)	0.469
Postoperative complications
Major morbidity	13 (14.3)	7 (13.7)	1.05 (0.39, 2.82)	0.927	11 (13.9)	7 (17.5)	0.76 (0.27, 2.15)	0.607	0.75 (0.26, 2.14)	0.588
Grade 3	6 (6.6)	4 (7.8)	0.83 (0.22, 3.09)	0.780	6 (7.6)	4 (10.0)	0.74 (0.20, 2.79)	0.655	0.74 (0.19, 2.81)	0.655
Grade 4	6 (6.6)	2 (3.9)	1.73 (0.34, 8.90)	0.508	4 (5.1)	2 (5.0)	1.01 (0.18, 5.78)	0.988	0.98 (0.16, 6.00)	0.984
Grade 5	1 (1.1)	1 (2.0)	0.56 (0.03, 9.08)	0.676	1 (1.3)	1 (2.5)	0.50 (0.03, 8.21)	0.621	0.44 (0.03, 7.69)	0.574
Length of stay, days
Mean (mean ± SD)	10.38 ± 11.79	9.45 ± 7.86	N/A	0.614	9.86 ± 11.20	9.20 ± 7.04	N/A	0.734	N/A	N/A
Median	7.0 (5.0-11.0)	6.0 (4.0-11.0)	N/A	0.411	7.0 (5.0-11.0)	6.5 (4.0-10.8)	N/A	0.717	N/A	N/A
≥ 14 days	16 (17.6)	12 (23.5)	0.69 (0.30, 1.61)	0.393	14 (17.7)	9 (22.5)	0.74 (0.29, 1.90)	0.533	0.71 (0.27, 1.85)	0.484
Discharge disposition			N/A	0.696			N/A	0.640	N/A	N/A
Home	81 (90.0)	46 (92.0)			72 (92.3)	35 (89.7)
Community hospital	9 (10.0)	4 (8.0)			6 (7.7)	4 (10.3)
Nursing home	0	0			0	0
30-day readmission	9 (9.9)	10 (19.6)	0.45 (0.17, 1.19)	0.103	8 (10.1)	7 (17.5)	0.53 (0.18, 1.59)	0.252	0.55 (0.18, 1.66)	0.292
Mortality
30-day	1 (1.1)	1 (2.0)	0.56 (0.03, 9.08)	0.676	1 (1.3)	1 (2.5)	0.50 (0.03, 8.21)	0.621	0.44 (0.03, 7.69)	0.574
1-year	6 (6.6)	3 (5.9)	1.13 (0.27, 4.72)	0.867	6 (7.6)	3 (7.5)	1.01 (0.24, 4.28)	0.985	1.04 (0.24, 4.55)	0.961
MBI at 6 weeks postoperatively	100 (98.5-100)	100 (100-100)	N/A	0.770	100 (99.0-100)	100 (98.0-100)	N/A	0.866	N/A	N/A
> 10% drop in MBI 6 weeks after surgery, yes	7 (7.7)	4 (7.8)	0.98 (0.27, 3.52)	0.974	6 (7.6)	3 (7.5)	1.01 (0.24, 4.28)	0.985	1.11 (0.25, 4.98)	0.892

¹Univariate analysis.

²Multivariate analysis.

Data are presented as median (interquartile range) or n (%). ICU: Intensive care unit; HDU: High dependency unit; MBI: Modified Barthel index; OR: Odds ratio; CI: Confidence interval; N/A: Not applicable; PSM: Propensity score matching; SD: Standard deviation.

DISCUSSION

ERAS has been shown to result in lower postoperative complications and shorten LOS following elective colorectal surgery[26]. Several studies have demonstrated that prehabilitation alone is associated with a reduction in postoperative complications and LOS in patients undergoing elective colorectal surgery[27,28]. These studies exhibited heterogeneity in both the prehabilitation protocols and patient selection criteria with several being observational in nature and lacking randomization. Our study aimed to evaluate whether refining the selection criteria, specifically targeting patients who were frail or with greater comorbidity burden, would better identify those who may benefit from prehabilitation. Through the use of PSM and covariate balancing, our study found that exercise-based prehabilitation did not confer significant postoperative benefits in patients who were not frail compared with those who did not undergo prehabilitation. We aimed to explore potential reasons underlying these findings.

Although numerous studies, including multiple meta-analyses, have demonstrated a reduction in overall complications among patients undergoing major abdominal surgery with prehabilitation[29,30], the 2022 clinical practice guidelines by ASCRS and SAGES only provided weak recommendations (level 2B evidence) supporting the use of prehabilitation in elective colorectal surgery: “Multimodal prehabilitation before elective colorectal surgery may be considered for patients with multiple comorbidities or significant deconditioning”[31]. The strength of the recommendation was limited by the heterogeneity in study populations (e.g., patient demographics, type of surgery, open or minimally invasive surgery), prehabilitation protocols (e.g., home-based vs day vs inpatient prehabilitation, duration of prehabilitation), and limitations in statistical power. This study sought to address these limitations by employing PSM to minimize the impact of confounding variables.

In light of the absence of a significant benefit from our prehabilitation program, important questions arise regarding the future role of prehabilitation in older patients undergoing elective colorectal surgery. Specifically, it remains to be determined whether prehabilitation should continue to be offered universally or selectively, targeting patients at higher risk for adverse outcomes as suggested by the 2022 ASCRS and SAGES guidelines[31]. Patients included in the non-prehabilitation group were patients who did not meet the criteria for prehabilitation (i.e. not frail, CCI < 4, and only required minimal assistance with activities of daily living) or refused prehabilitation for various reasons. Consequently, there is an inherent selection bias as patients in the prehabilitation group were generally more medically complex. It is plausible that the potential benefits of prehabilitation were attenuated or outweighed by the higher baseline burden of frailty and comorbidities in this group (notably, patients in the prehabilitation group had significantly higher incidence of frailty and CCI ≥ 4). This hypothesis was similarly proposed in a previous study evaluating a home-based prehabilitation program compared to standard care prior to elective pancreaticoduodenectomy[32]. Hence, PSM was performed to adjust for potential confounding covariates.

It is important to highlight that following PSM there were only 25 patients (17.4%) who were frail. Ideally, a subgroup analysis comparing outcomes between frail patients who received prehabilitation and those who did not would best determine the effectiveness of prehabilitation in this high-risk group. However, due to the small sample size and insufficient statistical power (as evidenced by our sample size calculation), we instead conducted a subgroup analysis among non-frail patients. Our analysis revealed no statistically significant differences between prehabilitation and no prehabilitation in terms of major morbidity, 30-day mortality, 1-year mortality, LOS, and > 10% decline in MBI at 6 weeks postoperatively. These findings suggest that exercise-based prehabilitation may offer limited benefits to patients who despite having multiple comorbidities (defined as CCI ≥ 4) are not frail.

Furthermore, the overall comorbidity burden as reflected by the CCI may not be as predictive of postoperative outcomes compared to the severity of specific comorbid conditions. While the CCI is a well-validated and comprehensive tool that incorporates age, chronic obstructive pulmonary disease, diabetes mellitus, and even dementia, existing literature suggests that certain comorbidities, such as atrial fibrillation and chronic obstructive pulmonary disease, are more strongly linked with postoperative complications than others, such as dementia[33]. Additionally, old age alone (≥ 80 years gives a CCI of 4) has not been consistently shown to predict increased morbidity or mortality[34]. These findings underscore the need for further studies to identify the group of patients who will benefit from prehabilitation the most and potentially shift focus from overall comorbidity burden to frailty and the severity of specific comorbidities.

Moreover, the absence of statistically significant findings in our study may be attributed in part to its retrospective design. A more robust approach to addressing confounding factors would be the implementation of a randomized controlled trial (RCT) that allows for balanced baseline characteristics and elimination of selection bias through randomization and standardization of both the intervention and control arms[35]. For instance, the recently published PREHAB RCT by Molenaar et al[28] that included 251 patients showed a reduction in medical complications and severe complications (defined as comprehensive complication index > 20) when compared to standard ERAS. The intervention arm underwent a standardized, multimodal 4-week in-hospital prehabilitation program, which included high-intensity physical training, nutritional optimization, psychological support, and smoking cessation (when applicable). However, the observed improvements may be due to the synergistic effect of these combined modalities rather than physiotherapy alone.

This hypothesis is further supported by a systematic review conducted by Bruns et al[36] (n = 3 RCT, 2 case-control studies) that evaluated the impact of exercise-based prehabilitation; none of the included studies demonstrated significant differences in postoperative complications and LOS[37-41]. Our study likewise assessed the impact of an exercise-based prehabilitation program in older patients undergoing elective colorectal surgery. Notably, all patients regardless of group allocation received the standard perioperative care, which includes the ERAS protocol, nutritional assessment and optimization by a dietician, counselling by a specialized nurse clinician, and preoperative and postoperative geriatric reviews. The improvement in postoperative complications reported by the PREHAB RCT may be attributed to combined effect of physiotherapy, nutritional, and psychological interventions rather than an exercise-based intervention alone[28]. Malnutrition is an established risk factor for anastomotic leak and other postoperative complications such as pneumonia[42]. This highlights the importance of a comprehensive, multimodal approach to prehabilitation.

The modality of prehabilitation (in-hospital, day-based, or outpatient) may also influence postoperative outcomes. In-hospital prehabilitation offer the advantage of regular and supervised reviews by physiotherapists[28] along with continuous monitoring and support from nursing staff (who may also assist patients in mobilization); nurse-led mobilization has shown to be feasible and effective following elective colorectal surgery and total joint arthroplasty, although its application in the preoperative period has not been formally studied[43,44]. In contrast, home-based prehabilitation programs are subject to greater variability and depend heavily on patient motivation and adherence to prescribed prehabilitation exercises. The absence of supervision increases the risk of incorrect execution of prescribed exercises, potentially leading to suboptimal training of targeted muscle groups and diminishing the therapeutic benefit[32]. Despite these limitations there is evidence to support the efficacy of home-based prehabilitation. An RCT by Bausys et al[45] (n = 122 patients) compared a semi-supervised home-based prehabilitation program with standard care in patients undergoing surgery for gastric cancer, showing a 60% reduction (relative risk = 0.40, 95%CI: 0.24-0.66) in incidence of 90-day postoperative complications. To date no studies have directly compared in-hospital vs home-based prehabilitation programs. However, a study on postoperative rehabilitation following total knee arthroplasty reported no difference in functional outcomes between an inpatient vs a monitored home-based program[46]. Whether in-hospital prehabilitation yields superior functional or postoperative outcomes compared with home-based prehabilitation remains a question to be answered.

Monitoring prehabilitation goals is critical given the variability in patients’ baseline functional status and comorbidity profiles. Objective markers should be used to guide individualized and progressive adjustments to prehabilitation interventions. In our study we used easy-to-administer assessments such as the MBI and 10 m walk test. While useful, these are broad indicators of functional performance and may lack sensitivity to detect nuanced changes in physiological reserve. The cardiopulmonary exercise testing (CPET), which evaluates parameters such as peak oxygen uptake, anaerobic threshold, and ventilatory efficiency, is considered the gold standard for assessing cardiopulmonary and musculoskeletal physiological functions[47]. CPET has demonstrated high sensitivity to physiological changes and is increasingly regarded as the preferred method for monitoring prehabilitation progress[48]. Although its implementation may pose logistical challenges, future studies evaluating the impact of prehabilitation on postoperative outcomes should consider incorporating CPET to enhance the accuracy and responsiveness of physiological assessments.

Our study possessed several notable strengths. It represented a decade-long investigation with over 200 patients included, providing a substantial dataset. PSM was also used to address the inherent selection bias associated with the retrospective study design and prehabilitation group allocation. Furthermore, all patients received care under a well-established perioperative GSS that has demonstrated consistently positive outcomes over the past decade[49].

However, several limitations should be acknowledged. This was a single-center study in an Asian population, and results may not be applicable to other centers. Despite the long duration of the study, the final sample size did not meet the calculated requirement of 359 patients required to detect a reduction in morbidity from 40% to 30%. Additionally, there was no standardized protocol regarding the type or duration of prehabilitation as these decisions were made collaboratively between the GSS team, the patient, and their family. Specific data on the nature of prehabilitation interventions were also not collected, and frailty-specific outcomes were not included. This reflects real-world challenges in which standardization is often hindered by resource limitations (e.g., limited availability of in-hospital beds), financial considerations (e.g., inpatient costs, travel expenses), and variable patient compliance with home-based prehabilitation. Moreover, the non-frail subgroup included patients who were classified as pre-frail (FFP of 1-2). While we used the FFP to define frailty, there are several frailty assessment tools available, each with a differing set of criteria. Nonetheless, FFP remains a validated and widely used measure with strong prognostic value for predicting postoperative morbidity and mortality[50].

CONCLUSION

Among older patients who were not frail, exercise-based prehabilitation did not demonstrate a significant reduction in postoperative complications or 1-year mortality following PSM. Open surgery was an independent predictor of postoperative morbidity and 1-year mortality. Further research on prehabilitation should consider refining the patient selection criteria by stratifying patients based on frailty scores and comorbidity burden in order to effectively identify those most likely to benefit, particularly in light of resource limitations in real-world clinical settings. Additionally, prospective studies comparing different modalities of prehabilitation (home-based vs day vs in-hospital) and their impact on postoperative outcomes should also be considered.

ACKNOWLEDGEMENTS

We would like to thank the Department of General Surgery of Khoo Teck Puat Hospital, as well as all members involved in geriatric surgery for the contribution towards the patients’ care.