Prehabilitation in elective abdominal cancer surgery in older patients: systematic review and meta‐analysis

Background Prehabilitation has emerged as a strategy to prepare patients for elective abdominal cancer surgery with documented improvements in postoperative outcomes. The aim of this study was to assess the evidence for prehabilitation interventions of relevance to the older adult. Methods Systematic searches were conducted using MEDLINE, Web of Science, Scopus, CINAHL and PsychINFO. Studies of preoperative intervention (prehabilitation) in patients undergoing abdominal cancer surgery reporting postoperative outcomes were included. Age limits were not set as preliminary searches revealed this would be too restrictive. Articles were screened and selected based on PRISMA guidelines, and assessment of bias was performed. Qualitative, quantitative and meta‐analyses of data were conducted as appropriate. Results Thirty‐three studies (3962 patients) were included. Interventions included exercise, nutrition, psychological input, comprehensive geriatric assessment and optimization, smoking cessation and multimodal (two or more interventions). Nine studies purposely selected high‐risk, frail or older patients. Thirty studies were at moderate or high risk of bias. Ten studies individually reported benefits in complication rates, with meta‐analyses for overall complications demonstrating significant benefit: multimodal (risk difference −0·1 (95 per cent c.i. −0·18 to −0·02); P = 0·01, I 2 = 18 per cent) and nutrition (risk difference −0·18 (−0·26 to −0·10); P < 0·001, I 2 = 0 per cent). Seven studies reported reductions in length of hospital stay, with no differences on meta‐analysis. Conclusion The conclusions of this review are limited by the quality of the included studies, and the heterogeneity of interventions and outcome measures reported. Exercise, nutritional and multimodal prehabilitation may reduce morbidity after abdominal surgery, but data specific to older patients are sparse.


Introduction
The majority of cancers in the UK are diagnosed in the older adult population (aged 65 years and above), with this population predicted to increase exponentially 1 . The pathogenesis and treatment of cancer can lead to a decline in cardiorespiratory fitness, weight loss and psychological morbidity 2 . Surgery remains the mainstay of curative treatment for many gastrointestinal, gynaecological and urological cancers, but outcomes are poorer in the older adult, making strategies to optimize this complex group increasingly important.
Adverse factors associated with ageing include co-morbidity, polypharmacy, cognitive impairment, dependency and frailty, all of which are associated with increased all-cause mortality in the general population 3 . When these at-risk individuals are exposed to the stress of major cancer surgery, postoperative mortality and morbidity also increase 4,5 . Common lifestyle choices, including smoking, poor nutrition and sedentary behaviours, add to this risk. 'Prehabilitation', the process of enhancing an individual's functional capacity before elective surgery with the aim of improving tolerance to the anticipated physiological stress of major surgery, may have a role in improving postoperative outcomes 6 . Prehabilitation programmes vary in their components, but can include exercise programmes, nutritional or psychological interventions 7 . Where they encompass different types of intervention, they are referred to as 'multimodal' 8 . In the context of the older adult, programmes may also include preoperative comprehensive geriatric assessment (CGA) and optimization. A summary of intervention types is presented in Fig. 1.
Early prehabilitation studies focused on the safety and feasibility of unimodality interventions 9 . More recently, studies have been more likely to be multimodal and to involve higher-risk populations 10 . Previous systematic reviews [11][12][13][14][15][16][17][18][19][20] focused predominantly on single-modality prehabilitation in mixed surgical populations. This review addresses the need for an updated review of the entire spectrum of prehabilitation interventions in elective abdominal cancer surgery with particular relevance to the older patient.

Methods
This systematic review and meta-analysis was conducted with reference to the Cochrane Handbook and is reported using the PRISMA guidelines 21 . The protocol was registered with PROSPERO (CRD42019120381). The primary objective was to determine whether any modality of prehabilitation (alone or in combination) before elective abdominal surgery leads to a reduction in either length of hospital stay (LOS) or complications (overall, pulmonary, wound infection rate, delirium, severe complications) compared with a control arm that does not include prehabilitation. The review was undertaken with particular relevance to older adults. Secondary objectives were to determine any effect on functional outcome measures (physical activity or walking capacity, weight loss, discharge independence) and psychological outcome measures (quality of life (QoL)). set. Searches were limited to studies published in the English language as resources were not available to support translation. The search was constructed using the PICO (patient, intervention, comparison, outcome) framework: Patient (adults undergoing abdominal or gastrointestinal surgery); Intervention (prehabilitation or preoperative optimization); Comparator (standard care or rehabilitation only); and Outcome (primary: LOS or complication rates). Clinical.Trials.gov was also searched for trials that had been completed but not published. A sample search strategy is shown in Appendix S1 (supporting information).

Inclusion and exclusion criteria
Randomized, case-control, cohort or retrospective studies reporting on adults (aged 18 years or above) undergoing surgery with curative intent for any gastrointestinal (oesophagus, stomach, pancreas, liver, colorectal) or intra-abdominal (urological or gynaecological) cancer were included. Studies including mixed surgical populations were included if they reported the cancer and non-cancer results separately or if more than 50 per cent of the population were patients with cancer. Studies could test any prehabilitation intervention or preoperative optimization strategy, alone or in combination (multimodal), and had to report outcomes in a control group. Control groups could include standard care, placebo, postoperative rehabilitation programme only, information leaflet or verbal advice on preparing for surgery and positive behaviour change (for example smoking cessation or alcohol reduction) in line with current perioperative care guidelines. Studies of postoperative interventions only were excluded, as were studies that did not report on either of the primary outcomes. Studies published only in abstract form without full text were excluded. Reference lists of primary studies and relevant systematic reviews were also hand-searched for additional studies. Screening of all titles and abstracts was undertaken independently by two reviewers. Articles were considered for full-text review if they met the study inclusion criteria or could not be excluded on the basis of the abstract alone.
Full-text articles were retrieved and assessed by the same two reviewers. Disagreements were addressed by discussion and consensus and, if required the opinion of a third reviewer was sought.

Definitions of eligible interventions
Eligible interventions included exercise interventions (either alone or in combination with pulmonary exercises), nutritional assessment and supplementation, psychological interventions, CGA and optimization, smoking cessation and multimodal (two or more modalities). These are summarized in Fig. 1. and other source of bias. Non-randomized studies were assessed on bias due to confounding, selection, classification of interventions, deviations from intended interventions, missing data, outcome measurement and reporting. Quality assessment was undertaken independently by two reviewers, and disagreements were resolved by consensus.

Data extraction
Data were extracted according to a predesigned pro forma, which included study characteristics, baseline data, intervention characteristics, adherence and outcomes. Studies were divided according to modality: exercise (alone or including pulmonary training), multimodal, nutrition, psychological, smoking, and CGA with optimization. The primary outcomes, LOS and complication rates, were recorded as mean(s.d.) values and proportions respectively. Where the mean was not reported, an approximation was calculated from the median and range 22 . Complication rates were recorded as total, severe (Clavien-Dindo grade III or above) or pulmonary complications, wound infections and delirium within 30 days of surgery. Secondary outcomes were extracted where reported: change in functional outcome measures (preoperative change in 6-minute walk test (6MWT) or cardiopulmonary exercise test (CPET) variables of physiological fitness, percentage preoperative weight loss or discharge independence), or psychological outcomes (postoperative Hospital Anxi-

Statistical analysis
Qualitative analyses were performed for all studies that met the inclusion criteria. Studies were analysed according to the type of prehabilitation intervention. Meta-analysis was performed using RevMan software (Review Manager version 5.3, 2014; The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark)

Control Mean difference (days) Mean difference (days)
Barberan-Garcia et al. 10 Boden et al. 36 Dronkers et al. 42 Carli et al. 9 Dunne et al. 43 Banerjee et al. 25 Barberan-Garcia et al. 10 Boden et al. 36 Dronkers et al. 42 Dunne et al. 43 Soares et al. 45 Banerjee et al. 25 Barberan-Garcia et al. 10 Boden et al. 36 Dronkers et al. 42 Dunne et al. 43 Soares et al. 45 Santa  Mantel-Haenszel random-effects models were used for meta-analysis; risk differences are shown with 95 per cent confidence intervals. c An inverse-variance model was used for meta-analysis; mean differences are shown with 95 per cent confidence intervals.  Bousquet-Dion et al. 47 Gillis et al. 26 Kaibori et al. 28 Li et al. 49 Mazzola et al. 50 Minnella et al. 29 Souwer et al. Bousquet-Dion et al. 47 Chia et al. 48 Gillis et al. 26 Jensen et al. 27 Kaibori et al. 28 Li et al. 49 Mazzola et al. 50 Minnella et al. 29 Nakajima et al. 51 Souwer et al. where the number (greater than 3) and quality of studies permitted, if the 95 per cent c.i. overlapped and effect sizes were similar 24 . Meta-analysis was performed using random-effects models, assessing risk difference for both dichotomous and continuous outcomes. Heterogeneity was assessed using the I 2 statistic. Significance was set at α = 0⋅050.

Results
Searches were performed on 6 January 2019. Some 130 papers were identified for full text review; 97 were excluded, leaving 33 studies for inclusion (Fig. 2).

Baseline characteristics
The studies, published between 2000 and 2019, included 2028 patients undergoing prehabilitation and 1934 controls. Interventions comprised: exercise only (9

Fig. 5 Forest plot showing the effect of nutrition prehabilitation on overall complications
Burden et al. 40 Kabata et al. 31 Gillis et al. 30 Kong et al. 41 MacFie et al. 32 39 . Sample sizes ranged from 32 to 443 patients, with most having fewer than 100 patients in each arm; only four studies 36,37,54,55 had more than this, and were mostly non-randomized. The wide range of sample sizes reflects the diverse primary outcomes on which power calculations were based, and also the fact that a small number were pilot or feasibility studies. Studies were predominantly single-centre, with only eight studies 33,[36][37][38]40,44,45,53 conducted across multiple centres. Studies were conducted in North America, Europe, Australasia, South-East Asia and Brazil. A range of surgical populations were studied, including colorectal (16 studies), upper gastrointestinal, hepatobiliary and pancreatic (9 studies), urological (3 studies), and mixed populations of gastrointestinal and abdominal malignancies (5 studies) (Table S1, supporting information).
Twenty-four studies involved patients with cancer exclusively, with a range of 52-78 per cent of patients with cancer in the remaining studies. Six studies included patients receiving neoadjuvant therapy. Although the average age range was 55-81 years, it was less than 70 years in the majority of studies. Three 48,50,52 of the ten multimodal studies and four 37,38,54,55 of the CGA studies had populations with an average age over 75 years (Table S1, supporting information). Nine studies 10,37,38,42,48,50,52,54,55 selected patients who were either assessed as frail (using a recognized frailty screen or criteria) or over a certain age cut-off; however the method of detecting frailty, frailty criteria used, and age varied between studies. Two studies 40,41 selected patients who were malnourished, and one 28 selected patients with chronic liver injury (Table S1, supporting information).

Methodological quality assessment
The assessment of methodological quality is summarized in Tables 1 and 2. Only three randomized studies blinded both participants and researchers, one 30 by using a placebo oral nutritional supplement, the second 36 by having all patients attend a preoperative physiotherapy appointment in which those in the control arm received only an information booklet whereas patients in the intervention arm learned breathing exercises, and the third 10 by using a double-informed consent model where control and intervention arms were not aware of each other. The absence of blinding of either participants or study personnel was the most common reason for high risk of bias assessment. The majority of RCTs adequately described randomization, but allocation concealment was not as reported robustly. Half of the RCTs adequately described blinding of outcome assessment 10,25,26,29,30,[34][35][36][38][39][40]42,43 . Only two studies 27,34 did not adequately report their outcome data ( Table 1).
Seven 49-55 of the eight non-randomized studies were graded as moderate risk of bias owing to bias in outcome measurements and due to confounding factors as they mainly used historical controls. One study 48 was judged to be at high risk of bias as the authors chose to include a wider age range in the intervention group than in controls ( Table 2).

Exercise-based interventions
Unimodal exercise interventions were most commonly based in hospital and conducted under supervision 36,42,43,45,46 ; four studies 36,42,45,46 included specific pulmonary exercises or training. Exercise prehabilitation programmes varied in intensity from a single preoperative session 36 to one to three times per week, and ranged from 1 to 6 weeks in duration.

Multimodal interventions
Multimodal interventions were more likely to be home-based 26,29,[49][50][51] ; all included exercise and nutrition, with four 26,47,49,52 also including psychological interventions. The nutritional component of multimodal interventions commonly involved dietician assessment and supplementation if required. Two studies 28,48 did not mention supplementation. Two multimodal programmes specifically mentioned other behavioural modifications: alcohol reduction 49 and smoking cessation 50 .

Nutrition-based interventions
All nutrition-only prehabilitation studies [30][31][32][33]40,41,53 included oral nutritional supplementation, but the prescriptions varied from 'ad libitum' between meals to 400 ml three times a day, with duration varying from 1 to 4 weeks. Two studies 32,33 included separate intervention groups that received supplements both before and after surgery.

Psychology-based interventions
The two psychological prehabilitation studies had different interventions; the study by Chaudhri and colleagues 34 looked at the impact of a community-based stoma education intervention, whereas that by Haase and colleagues 35 involved giving patients audio recordings with either guided imagery or relaxation techniques to listen to before surgery.

Comprehensive geriatric assessment with optimization
All four CGA prehabilitation studies 37,38,54,55 involved preoperative CGA performed by a geriatrician-led multidisciplinary team, nutritional optimization and medication reviews; two studies 37,54 included postoperative daily reviews by a geriatric specialist nurse. Two studies specified that they corrected anaemia with either blood transfusion 55 or supplementation 38 .

Smoking cessation
One study 39 of a smoking cessation intervention met the inclusion criteria; the intervention involved a single smoking cessation counselling session combined with nicotine replacement therapy.

Adherence
Adherence was reported in eight 9,10,25,36,[42][43][44]46 of the nine studies of exercise, five 26,27,29,47,49 of the ten multimodal studies, and four 30,32,40,41 of the seven nutrition prehabilitation studies, with percentages varying from 69 to 100 per cent, 59 to 98 per cent, and 75 to 99 per cent respectively. Adherence was not stated in studies of psychological, CGA with optimization, or smoking cessation interventions; as these were typically single preoperative interventions, adherence would not have been an issue.

Primary outcome
Twenty different primary outcomes were reported, and 12 of the 33 studies reported more than one primary outcome measure (Tables 3-8). Four studies 25,27,42,44 reported feasibility as the primary outcome. Postoperative complications (overall complication rate, severe complications (Clavien-Dindo grade II or above, or III or above), pulmonary complications, delirium or site-specific infection rate) were the most common postoperative outcome measures, and were reported in all except one study 34 . LOS was reported in all except two studies 31,46 .

Exercise studies
One study 10 reported a significant reduction in overall complications in the intervention arm (20 of 62 versus 38 of 63 in the control arm, P = 0⋅001; relative risk 0⋅5 m, 95 per cent c.i. 0⋅3 to 0⋅8). One study 9 found a non-significant higher overall complication rate in the intervention arm (22 of 56 versus 18 of 54 for the control; P value not reported), which was attributed to poor compliance in the intervention group and an increase in physical activity in the control group. Meta-analysis showed no significant difference in overall complications, but heterogeneity was high (Fig. 3a).
Two studies reported lower rates of pulmonary complications in the intervention group: 27 of 218 versus 58 of 214 (adjusted hazard ratio 0⋅48, 95 per cent c.i. 0⋅30 to 0⋅75; P = 0⋅001) in the study by Boden and colleagues 36 , and five of 16 versus 11 of 16 (P = 0⋅03) in that of Soares and co-workers 45 . Yamana et al. 46 also found a lower Clavien-Dindo grade of pulmonary complication with intervention (P = 0⋅014). Meta-analysis of five studies (the study by Boden and colleagues 36 was excluded owing to a significantly different intervention) for pulmonary complications revealed a non-significant trend in favour of the intervention (Fig. 3b).
A non-significant trend towards lower LOS was also observed on meta-analysis ( Fig. 3c and Table 3).
Two studies 25,43 that assessed preoperative change in CPET variables before and after intervention both demonstrated significant improvements in peak oxygen uptake and peak work rate ( Table 3). Four studies 9,10,44,45 that assessed functional walking ability using the 6MWT demonstrated no preoperative differences between intervention and control groups. Of the five studies that reported psychological outcomes, only that by Dunne and colleagues 43 showed an improvement in overall QoL score measured using the SF-36 ® (+11, 95 per cent c.i. 1 to 21; P = 0⋅028) and overall mental health score (+11, 1 to 22; P = 0⋅037) ( Table 3).

Multimodal studies
One study 50 found a reduction in overall complications in the intervention group (17 of 41 versus 26 of 35 in the control group; P = 0⋅005) ( Table 4). Meta-analysis showed a significant reduction in overall complications after multimodal prehabilitation (Fig. 4a). Mazzola Table 4). Meta-analysis for LOS including six studies was not significant; however, there were high levels of heterogeneity (Fig. 4b).
Four multimodal studies 26,29,49,51 demonstrated significant preoperative improvements in functional walking ability using the 6MWT after the intervention (mean difference range 24-62 m; all P < 0⋅010) ( Table 4). However, in two of these studies 49,51 walking ability was tested only in the intervention group. No differences in psychological outcomes were observed in multimodal studies 47,49,59 ( Table 4).

Nutrition studies
Two studies reported a reduction in overall complications in the intervention group: eight of 54 versus 17 of 48 in the control group (P = 0⋅04) in the study by Kabata and colleagues 31 , and 15 of 32 versus 34 of 44 respectively (P < 0⋅050) for group 2 in the study by Smedley et al. 33 (Table 5). Meta-analysis demonstrated significantly fewer overall complications following the intervention (the historical study of MacFie et al. 32 was excluded from meta-analysis) (Fig. 5).

Psychological studies
Chaudhri and co-workers 34 reported a reduction in LOS in the intervention group (8 versus 10 days in the control group; P = 0⋅029), which was attributed to fewer delayed discharges owing to stoma proficiency ( Table 6). Haase et al. 35 found no difference in overall complications between either of their interventions and the control. Neither psychological intervention had any effect on the measured psychological outcomes 34,35 (Table 6).

Smoking studies
The smoking cessation trial 39 did not find a reduction in either complications or LOS with intervention ( Table 8).

Discussion
This systematic review has found evidence from a number of trials that exercise, multimodal, nutrition and CGA with optimization prehabilitation programmes may reduce the number of postoperative complications after elective surgery for gastrointestinal and urological cancers. It has shown evidence that multimodal, nutritional, psychological and CGA interventions (but not exercise interventions or smoking cessation alone) may reduce LOS. In particular, the small number of studies that selected high-risk, frail or older patients were more likely to report improvements in either complications or LOS compared with studies that included all patients. Equally, studies conducted in patients undergoing oesophageal and upper gastrointestinal surgery, known to be associated with high levels of postoperative morbidity and mortality, were more likely to demonstrate reductions in pulmonary complications. However, conclusions are limited by the methodological quality of included studies, in particular the lack of blinding of participants in all except three studies. Significant heterogeneity of interventions also limits comparison. Adherence to exercise, multimodal and nutritional interventions was generally high; however, it is possible that participant selection bias and lack of blinding may have resulted in more motivated patients being recruited.
National and international guidelines 57-59 recommend that CGA should be performed in all patients over the age of 70 years with a diagnosis of cancer to try to predict treatment toxicity and postoperative complications, and to aid in shared decision-making. However, there remain very few studies of CGA in surgical cancer populations, and the majority of these are limited to its role in risk prediction and prognostication 60,61 . This systematic review identified only two RCTs 37,38 evaluating CGA and tailored interventions. It is worth noting that the median age of patients in studies included in this review was only 68 years, with patients in the exercise-alone interventions having a median age of only 63 years. Only seven of the 33 studies in this review had a median age greater than 75 years. This suggests that many prehabilitation studies to date either failed to recruit older patients due to the location or nature of the interventions or they excluded older patients owing to a perceived risk of the interventions, despite mounting evidence 62,63 that exercise-based interventions are safe in older individuals.
This review also demonstrated that improvements in preoperative functional measures can be made with exercise prehabilitation (measured by CPET), multimodal interventions (measured using 6MWT) and nutritional prehabilitation (reduction in preoperative weight loss). However, the link between small statistically significant improvements in these variables and clinical outcomes is not clear.
A number of previous systematic reviews have examined individual components of prehabilitation in varying surgical populations: exercise [18][19][20]64,65 , exercise in frail individuals 16 , multimodal interventions [13][14][15] , multimodal interventions in frail individuals 12 , nutrition with and without exercise 66 , and psychological interventions 11 . All of these, including the present review, have been limited by the quality of the underlying evidence. This is the first review that included all modalities of prehabilitation of relevance to the older adult.
Prehabilitation programmes, regardless of the individual components they comprise, are complex multicomponent interventions, and thus should be evaluated as such. The Medical Research Council in the UK has published a clear framework for evaluating and conducting trials involving complex interventions 67 . Two of the potential reasons for negative findings in prehabilitation studies are either that the interventions are too standardized to enable reproducible delivery or that, in efforts to provide truly personalized programmes, no two individuals receive the same intervention. Equally, although there is accumulating evidence that multimodal prehabilitation is likely to be more beneficial than using a single modality, future trials that use methodologies designed for evaluating complex interventions will be able to determine which components are most beneficial for different patients and why. This review is limited by the heterogeneity of outcomes reported. LOS and complications were selected as primary outcomes for this review; however, a number of studies were powered to detect changes in other primary outcomes and therefore may have been inadequately powered for the primary outcomes of this review. The majority of trials in prehabilitation are relatively small, and this may contribute towards reporting bias of trials with statistically significant outcomes. Heterogeneity of studies may have also contributed to some analyses attaining statistical significance inappropriately. The wide date range of included studies may have added to the heterogeneity, as perioperative care has evolved over the past 20 years with the introduction of enhanced recovery pathways and laparoscopic surgery. Another potential limitation is that diverse surgical procedures with a range of complication rates have been compared. This may have resulted in some analyses not reaching significance, and will have contributed towards heterogeneity on meta-analysis. For the purpose of this review, a large number of studies were excluded at full-text review due to lack of reporting of LOS or complications, which are considered core outcomes for surgical trials 68,69 . In particular, a number of trials of psychological interventions [70][71][72][73][74][75] were excluded for this reason. Of note, only one preoperative smoking cessation trial 39 and no studies in gynaecological cancer surgery met the inclusion criteria. The main strength of this review is the comprehensive nature, whereby all current prehabilitation modalities in abdominal cancer surgery were included. This means that the review is of relevance to a wide range of surgical specialties, identifies gaps in the current evidence base, and will be of interest to commissioners looking to fund prehabilitation services.
The reporting of outcomes presented a challenge in this review owing to the range of outcome measures used; this reflects complex interventions and the inability to compare them directly, and raises an important issue for researchers. The evidence base for prehabilitation might be stronger if a core outcome set could be used in all trials, irrespective of modality of prehabilitation or surgical population, to facilitate comparison of interventions. The StEP-COMPAC group (Standardising Endpoints in Perioperative Medicine) have already made progress in this regard in perioperative medicine [76][77][78][79] . Initiatives such as the DiSCO (Defining Standards in Colorectal Optimisation) project led by researchers in the West of Scotland, which aims to create key sets of standards for prehabilitation in collaboration with patients, their caregivers and the public, will be vital in ensuring that results are relevant to service users as well as clinicians, and to the successful promotion of patient-centred care. Future studies also need to evaluate strategies for implementation and the associated costs to enable adequate investment at a time of increasing healthcare costs.