We conducted this systematic review according to a previously published protocol (He et al., 2022), that was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO) on October 10, 2021, as CRD42021277191. This systematic review was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement (Page et al., 2021), and the guidelines of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., 2022).

As described in our protocol (He et al., 2022), studies were included in the review if they met the following inclusion criteria: (a) Participants–participants aged 60 years and older who were scheduled for elective surgery of any type, with no limits to anaesthetic technique, depth, or agents, regardless of race or sex; (b) Intervention and comparison–cognitive prehabilitation (preoperative cognitive interventions, such as cognitive training, stimulation, and rehabilitation) compared to standard or usual care, or no cognitive prehabilitation under the same perioperative treatment programs; the delivery of preoperative cognitive interventions was not limited to a specific mode (paper-and-pencil, computer-administered, individual, or group-based), and there were no restrictions with respect to setting (hospital, community, or home-based) or intervention dose-related parameters, including the overall duration of the intervention, frequency, and intensity; however, we excluded single-session treatments; (c) Outcome measures—the primary outcomes were the incidence of POD and the incidence of delayed neurocognitive recovery (dNCR); the secondary outcome measures included any validated measure assessing cognitive function, including but not limited to, cognitive screening instruments, such as the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA), multidomain cognitive assessment scales, neuropsychological test batteries, postsurgical complications, activities of daily living (ADL), health-related quality of life (QOL), length of stay (LOS), mortality (30 days or longer if reported), patient compliance and acceptability, and safety (dropouts, serious adverse events); and (d) Study design and language—RCTs published in English or Chinese. Pilot, multiarm, and cluster RCTs were also eligible. We excluded studies that were research protocols, conference proceedings or abstracts, dissertations, or books as well as studies that lacked available data for analysis.

The MEDLINE (via Ovid), Embase (via Ovid), Cochrane Central Register of Controlled Trials (CENTRAL) (via Ovid), Cumulative Index to Nursing and Allied Health Literature (CINAHL) plus (via EBSCOhost), PsycINFO (via EBSCOhost), Physiotherapy Evidence Database (PEDro), Chinese Biomedical Literature Database (CBM), China National Knowledge Infrastructure Library (CNKI), WANFANG Database, and Chinese Scientific Journal Database (VIP) databases were systematically searched from inception to June 10, 2022, and the search was updated on September 5, 2024. The Patient, Intervention, Comparison, Outcome, and Study design (PICOS) framework was used to construct the search strategy: patient (aged), intervention (prehabilitation), outcome (PND) and study design (RCT) (He et al., 2022). Furthermore, we examined the reference lists of the included study reports and any relevant systematic reviews to identify additional eligible studies. The search strategy is shown in Supplementary Table S1.

The retrieved records were imported into EndNote (EndNote Citation Software, Version 9.3, Clarivate Analytics, New York, NY, United States), and duplicate items were removed automatically and manually checked by one author. Two authors independently inspected the titles, abstracts, and keywords of all the articles to eliminate ineligible records. After a preliminary screening, the full texts of the remaining records were retrieved. Two authors independently assessed the full texts in detail according to the predefined criteria and identified studies for inclusion. Discrepancies regarding the eligibility of the selected studies were resolved through discussion or consultation with a third author, if necessary.

Two authors independently extracted the relevant general information, participant and intervention characteristics, outcome data, and methodological quality components from the included studies using a predetermined data extraction Excel format. The PEDro scale (Maher et al., 2003) has a possible score range of 0–10, with higher scores suggesting higher quality; it was adopted as a criterion to quantify the methodological quality of each included RCT. The scores of each RCT were independently determined by two authors by evaluating the following criteria: randomization, allocation concealment, baseline comparability, blinding, follow-up, intention-to-treat (ITT) analysis, between-group statistical comparisons, and reporting point measures and measures of variability. Discrepancies regarding data extraction and quality assessments were resolved through discussion or consultation with a third author, if necessary.

We planned to conduct a quantitative synthesis by means of a meta-analysis. However, given the nature of the extracted data, this was deemed unsuitable to yield a meaningful summary effect estimate owing to the small number of included studies and the clinical and methodological diversity between the trials. Therefore, the available quantitative data were analysed narratively and synthesised in accordance with the Synthesis Without Meta-analysis (SWiM) guidelines (Campbell et al., 2020). The SWiM guidelines include a nine-item checklist designed to promote transparent reporting of systematic reviews that employ narrative synthesis without meta-analysis, which prompts users to report how studies are grouped, any standardised metric used, the synthesis method, how data are presented, a summary of the synthesis findings, and limitations of the synthesis (Campbell et al., 2020). The results of the included studies were summarised and presented through tabulation and description. In addition, we transparently reported, when available, the study outcomes of each trial through structured reporting of effects and calculated effect sizes with 95% confidence intervals (CIs) using RevMan 5 software. We presented the risk ratio (RR) with a 95% CI for dichotomous outcomes and the mean difference (MD) with a 95% CI for continuous outcomes. We applied the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework to assess the certainty of evidence. The GRADE framework includes five domains: study limitations, indirectness, imprecision, inconsistency and the likelihood of publication bias. Ultimately, the quality of evidence for each outcome falls into one of four levels: high, moderate, low, and very low (Guyatt et al., 2008; Murad et al., 2017). The results are presented in a narrative “Summary of findings” table (Supplementary Table S2).

In total, 3,734 records were identified from the database searches, which were reduced to 3,044 after duplicates were removed. After screening titles, abstracts, and keywords, 76 potentially relevant full papers were assessed, with six eligible studies included in this review (Vlisides et al., 2019; Saleh et al., 2015; Humeidan et al., 2021; Li et al., 2018; Wang and Wang, 2017; Greaves et al., 2023) The study selection process is documented in the PRISMA flow diagram (Figure 1). The characteristics of the excluded studies are presented in Supplementary Table S3.

The eligible studies were RCTs that used parallel-arm controls. The studies included 645 older adults, with 316 in the intervention group and 329 in the comparator group. The sample sizes in the included studies ranged from 29 (Greaves et al., 2023) to 251 (Humeidan et al., 2021) participants. The studies were published between 2015 and 2023. Three studies were conducted in China, two in the USA and one in Australia. Of the six included studies, four were published in English and two in Chinese. Further details concerning the characteristics of the included studies are presented in Tables 1 –3, ordered by study number.

This systematic review identified a small and inconclusive evidence base for the effectiveness and safety of cognitive prehabilitation on postoperative cognitive and noncognitive outcomes in older adults undergoing elective surgery. The effects of preoperative cognitive training on reducing the incidence of dNCR, the incidence of POD and LOS, the incidence of postsurgical complications, as well as improving postoperative global cognitive function and ADLs are quite uncertain. The results of this systematic review should be interpreted with caution because of the limited number of trials and the low to very low certainty of evidence.

Cognitive interventions are diverse nonpharmacological therapies based upon the distinct theoretical constructs aimed at preventing decline, restoring reduced function, and compensating for impairment (Sikkes et al., 2021). In our study, all six included studies investigated the effect of cognitive training, which typically refers to guided practice on a set of standardised tasks targeting a specific cognitive domain or domain such as memory, attention, or problem solving, and is intended to benefit cognitive functions (Clare et al., 2003; Bahar-Fuchs et al., 2013). It falls within the scope of complex intervention and needs to be developed and evaluated under the guidance of frameworks such as the Medical Research Council complex intervention research framework (Craig et al., 2008; Skivington et al., 2021). While it has often been the focus of research with clinical populations, such as those with dementia and mild cognitive impairment, there is increasing evidence about its utility in preventing cognitive decline (Butler et al., 2018; Gates et al., 2019), with the possible mechanism built on the premise of cognitive reserve (Stern, 2012).

In our study, the content and structure of the preoperative cognitive training were clinically heterogeneous, with most targeting several cognitive domains simultaneously and the other one single cognitive domain, some delivered through paper-and-pencil and others via computer-based platforms. The settings at which the interventions were delivered were also diverse, with some delivered at supervised hospital settings and others at home without supervision. Notably, the actual total training time was predominantly less than 5 h, which falls short of the 10 h presumed to be the effective “dose” of cognitive training (Willis et al., 2006; Edwards et al., 2018). Due to the limited studies included, core/specific ingredients and dosing parameters, especially the minimal effective dosage prior to surgery and the most effective and/or feasible type of cognitive interventions, remain to be investigated in future studies.

Attrition is frequently a barrier to cognitive training trials, particularly in unsupervised, home-based and computer-based settings. Therefore, the interpretation of the actual effects of the interventions may be biased in the context of suboptimal compliance. Although expecting perfect adherence may not be realistic, there is a need for supervision or intervention of home-based training by utilising devices with cellular capabilities that may allow for real-time analysis, automated reminders, targeted coaching, and more customised training in future studies (O’Gara et al., 2020; Humeidan et al., 2021).

Although cognitive prehabilitation offers hope for preemptive neurocognitive optimisation, it also raises awareness of how challenging it may be to integrate it in the surgical trajectory. This is also true for its physical counterpart, the compliance rates of which vary from 16 to 97% (Bruns et al., 2016; Punnoose et al., 2023; Carli et al., 2020). The identification of support and barriers to adherence is an important step towards the development and implementation of prehabilitation (Ng et al., 2022). Surveys of surgical patients have shown that factors related to surgical prehabilitation program adherence include patient-centered programs with tailored interventions (Wynter-Blyth and Moorthy, 2017; Ferreira et al., 2018; Gurunathan et al., 2023), interpersonal and environmental motivators (Wynter-Blyth and Moorthy, 2017; Parker et al., 2019; Gillis et al., 2021; Cooper et al., 2022), patient empowerment, and understanding the importance of prehabilitation (Cooper et al., 2022; Porter Starr et al., 2023). In terms of cognitive components, investigations targeting attitudes and perceptions towards cognitive training in older adults indicated that support strategies (e.g., regular personalized feedback and supervision, assistant service) (Trenorden et al., 2022; Beishon et al., 2022; Chen et al., 2021), identification and management of lack of motivation, low mood, and anxiety and depression (Beishon et al., 2022), as well as incorporating meaningful social engagement (Stephan et al., 2024), contributed to positive participant experiences. Therefore, protocol adherence may be a key mediator of prehabilitation efficacy; it is crucial to understand and address this barrier in future research.

Notably, the diagnostic tools and rules for delirium and dNCR differed across the included studies, which may partially explain their heterogeneity. This was consistent with the findings of the study by Borchers et al. (2021). This finding again confirms the importance of using a more scientific approach to define the outcomes measured in clinical research. Future efforts to define standards for perioperative cognition research, including diagnostic criteria or rules as well as core outcome sets, especially those better aligned with the recent American Society of Anesthesiologists statement on Perioperative Brain Health and recommendations of the Nomenclature Consensus Working Group, are warranted (Evered et al., 2018; Evered et al., 2019; Borchers et al., 2021; Gargon, 2016). The appropriate clinical interpretation of changes on an outcome must consider not only statistical significance, but also whether the observed change is of genuine clinical value to patients (McGlothlin and Lewis, 2014; Embry and Piccirillo, 2020). MCID defines the smallest amount an outcome must change to be meaningful to patients that demonstrates a clinical benefit of an intervention (McGlothlin and Lewis, 2014; Harris et al., 2023). The MCID can be determined using consensus, anchor, and distribution-based methods (McGlothlin and Lewis, 2014). It should also be noted that very few of the included studies defined or mentioned an MCID, or reported data on psychological well-being, QOL, daily functioning, and adverse events, which are important outcomes in clinical decision-making.

The results of our study are similar to those of four recent reviews in related fields (Daksla et al., 2022; Volz et al., 2022; Jiang et al., 2023; Zhao et al., 2024; Bowden et al., 2023), although these reviews reported on heterogeneous populations (Daksla et al., 2022; Volz et al., 2022; Jiang et al., 2023; Zhao et al., 2024; Bowden et al., 2023), or with indications for oncological or gynaecological surgery (Daksla et al., 2022; Volz et al., 2022), or targeting on postoperative cognitive interventions (Jiang et al., 2023; Bowden et al., 2023). The findings revealed that preoperative cognitive training might be a useful addition to multimodal surgical prehabilitation in perioperative pathways. Comments for two of the included trials (Vlisides et al., 2019; Humeidan et al., 2021) also reflected that cognitive prehabilitation could be beneficial for older surgical patients (Goettel, 2019; Keshock, 2022). However, this question remains to be answered given the paucity of data and the limitations that prevent researchers from drawing firm conclusions. With increasing awareness of the importance of optimising perioperative brain health, the importance of cognitive prehabilitation for the growing surgical population of older adult needs to be further demonstrated by future studies (Culley and Crosby, 2015; Lenze et al., 2019; Kato and Solt, 2021).

This study has some limitations that hinder the interpretability of the results. First, the limited number of studies and the heterogeneity among them precluded us from performing a planned quantitative synthesis. Second, despite conducting systematic searches in multiple relevant databases, our searches were restricted to studies published in the English and Chinese (the authors’ native language), and we did not search for grey literature, as prior studies have shown that excluding unpublished studies, dissertations (Schmucker et al., 2017; Hartling et al., 2017), and non-English (Hartling et al., 2017; Dobrescu et al., 2021; Nussbaumer-Streit et al., 2020) publications would have little impact on the overall conclusions. However, this may have resulted in relevant studies being missed, and the resulting language and publication bias may compromise the comprehensiveness and timeliness of this study. Moreover, there were minor amendments to the published review protocol, and we did not report any meta-analyses due to clinical and methodological heterogeneity, which may have led to misleading results that lacked generalizability. The differences between the protocols and reviews are described in Supplementary Table S5.

First, fundamental and early-stage efforts are needed to develop or identify cognitive prehabilitation regimens with greater acceptance of the participants, followed by preliminary trials with the aim of feasibility testing, evaluation, and implementation of intervention strategies. Important questions should focus on setting, timing, type, duration, frequency, and intensity as well as the minimum effective dosage needed to exert clinically important benefits. These findings will serve as a foundation for future well-designed, executed, and reported large-scale, appropriately powered RCTs in this field (Lobo et al., 2023). In addition, future studies should address the issues raised in the present review. At present, cognitive prehabilitation studies mainly involve noncardiac, nonneurological surgery, and more information is needed for other surgical specialties. Preoperative patient characteristics, including cognitive status, education, comorbidities, and medications, can provide valuable information for outcome and risk stratification. Studies should also report compliance with intervention strategies so that outcomes can be pooled or stratified according to the degree of compliance. Future studies should be better powered to demonstrate the cost effectiveness of cognitive prehabilitation. It is also important to standardise validated cognitive outcome measures in future research to decrease between-study heterogeneity and enable pooled analysis. This is further highlighted by the importance of initiating core outcome measures for perioperative and anaesthetic care (Boney et al., 2022) and mapping these alongside standardised endpoints in perioperative medicine (Moonesinghe et al., 2019; Haller et al., 2019). Finally, trials with prolonged follow-up are required to understand both the short- and longer-term beneficial effects of cognitive prehabilitation.