Is patient-specific pre-operative preparation feasible in a clinical environment? A systematic review and meta-analysis

Abstract Technical difficulty of an operation is associated with patient and disease characteristics, indicating the necessity for surgeons to exercise patient-specific preparation. Such methods have been shown to be effective in the simulation suite, however, application in a real clinical environment has been sporadic. This systematic review attempts to answer if patient-specific preparation in challenging surgical procedures is feasible. A systematic review of OvidMedline, Embase and all Evidence Based Medicine review databases, was conducted in search of studies who described surgical rehearsals in all specialties. Following the application of defined inclusion and exclusion criteria relevant data were extracted and summarised. Descriptive synthesis was performed for all included studies and meta-analysis of data was applied when possible. Of fourty-nine studies included, thirty-seven were case-series, ten were non-randomised comparative trials and two randomised controlled trials. Accuracy of applied methods ranged from 66.7 to 100% and a good outcome was seen in 60–100% of operations. Meta-analysis of studies comparing rehearsals to real procedures (same patients) showed that simulated procedures were significantly faster than real ones (SMD = −1.56 [−2.19, −0.93] p < 0.00001) but were similar in other outcomes (fluoroscopy time: SMD = −0.1 [−0.63, 0.42] p = 0.7, fluoroscopy volume: SMD = −0.43[−0.97, 0.11], p = 0.12). Meta-analysis of studies comparing pre-operative rehearsals to standard treatment (two distinct groups of patients), demonstrated that real procedures were performed quicker if pre-operative rehearsal took place (SMD = −0.47 [−0.79, −0.16], P  = 0.003) but the immediate clinical outcome was similar for practiced and not practiced operations (SMD =0.03[−0.23, 0.29], p = 0.82). Current evidence suggests that patient-specific pre-operative preparation is feasible and safe and decreases operational time.


Introduction
Patient and disease-specific characteristics are associated with an operation's technical complexity [1][2][3][4][5][6]. Hence, a methodical preparation towards challenging surgical procedures, encompassing patient-specific features, may have a significant effect on improving both the quality and outcome of surgery. Emerging technologies, such as 3D image reconstruction (e.g. fusion of 2D-into 3D-images using computed tomography (CT) slides), and rapid prototyping (building a physical model using computer aided design (CAD) software, e.g. 3D printing) facilitate the effective and accurate construction of patient-specific anatomical models, surgical guides and implants [7][8][9][10], thus opening new avenues for pre-operative rehearsals.
Despite the increase in popularity within the last twenty years the feasibility, accuracy and clinical impact of patient-specific pre-operative preparations have not been well established [7,8,11], impeding their uptake into routine clinical practice. The primary aim of this review was to evaluate whether patient specific pre-operative preparation is feasible in the time-pressured environment of surgical specialties. Moreover, safety, accuracy and clinical impact were assessed.

Methods
This systematic review was performed according to the PRISMA guidelines [12]. A systematic review of OvidMedline (1946-July 31st 2015), Embase (1947-July 31st 2015) and all Evidence Based Medicine (EBM) review databases, including Cochrane databases, was conducted using the following terms: "patient", "specific", "simulat Ã ", "surg Ã ", "planning", "patient specific" and "rehearsal". The search strategy was validated with the identification of eight studies familiar to the authors and believed to be appropriate for inclusion. Review papers were searched by hand for additional studies appropriate for inclusion.

Inclusion criteria
Studies on patient-specific pre-operative rehearsals or planning followed by the performance of real-time operations were included in this review. The term "rehearsals" is used here to reflect physical re-enactment using virtual surgical tools (i.e. similar to the ones used in real surgery), whilst planning refers to viewing an anatomical model and deciding on a surgical strategy with or without the use of virtual tools that are dissimilar to the ones used in a real theatre. Only studies written in English were included.

Exclusion criteria
Studies reporting merely construction of anatomical models or simulators without subsequent real-time operations were excluded because the aim of the review was to assess clinical impact of the preparation process. Case reports, reviews, editorials, abstracts and conference proceedings were excluded from further analysis. Studies on peri-operative navigation without pre-operative preparation were also excluded. Also, studies exclusively on paediatric or foetal surgery were excluded.

Identification of relevant studies
Two authors independently assessed studies identified through the literature search for relevance, by reading the title and abstract. Full text was retrieved for all studies potentially meeting the eligibility criteria.

Data extraction
Data extraction from selected studies was done using a MicrosoftV R Excel spreadsheet, which included the following categories: authors, country, year of publication, anatomical model type, type of surgery, number of participants, rehearsal/planning process, number of cases pre-operative planning was followed in, clinical outcomes, other outcomes, cost and time requirements for preparation of pre-operative planning session. Data extraction was conducted independently by two authors and subsequently validated by a third. Descriptive synthesis was performed for all included studies, structured around the PICO (Participants, Intervention, Comparators, Outcome measure) approach [13].

Meta-analysis
For the purposes of the meta-analysis assumption of normal distribution of the data was made and the following mathematical equations were used to calculate Standard Deviation (SD) on different occasions (collaboration): i. SD ¼ IQR (Interquartile range) width/1.35 ii. SD ¼ ͱN (CI (Confidence Interval) upper limit -CI lower limit)/4.128 iii. SD ¼ SE (Standard Error)/ͱ(1/N E þ1/N c ), SE ¼ Mean Difference (MD)/t N: sample size N E : sample size of intervention group N c : sample size of control group t: ratio of difference in means to standard error of difference in means

Results
The literature search yielded 1779 records. Of these, 652 were duplicate studies and were therefore excluded. Titles and abstracts of the remaining 1127 studies were assessed for relevance, by two of the authors. Full text was acquired for 130 studies. From these, 45 matched the inclusion criteria. Manual search of references (from review studies identified during

Participants
The participants were patients due to undergo surgery. Whilst most preparation processes involved only the surgeon, in 2 studies the entire surgical team was involved aiming to achieve improved coordination [41,42].

Type and content of intervention
The methods for patient-specific preoperative preparation included: i) surgical planning; allowing the surgeon to establish the surgical approach and dissection sequence, but not including a complete physical rehearsal of the procedure e.g. inspection of anatomy with augmented reality environment platforms [37,43], or ii) surgical rehearsal (i.e. simulated surgery), where the surgeon had a variety of virtual surgical tools at their disposal and performed all the physical and mental processes one would in a real theatre environment e.g. performing surgery on a virtual reality simulator [41,[44][45][46].

Comparators
Two types of comparisons were used in the included studies: (i) results of rehearsals were compared to the ones of real operations (i.e. patients act as their own controls) [41,44,45] (ii) preoperative preparation methods were compared to standard treatment or computed aided surgery (i.e. two distinct groups of patients) [15,21,38,51]. The overall outcome in question was similarity between rehearsal and real procedure on the first occasion and assessment of possible superiority of the rehearsed procedures compared to non-rehearsed ones (evaluating for possible differences) in the second type of comparisons.

Type of outcomes
Descriptive synthesis of data regarding measured outcomes yielded the following commonly emerging themes: (i) accuracy of pre-operative preparation methods and (ii) clinical outcomes.

Clinical outcomes
Immediate surgical outcomes, including peri-operative complications, were reported in most studies. For comparison, the number of cases that had a satisfactory outcome (e.g. anatomical reduction of a fracture) are presented as a percentage of the overall number of cases ( Table 2).
The majority of studies reported satisfactory immediate surgical results (60-100%). Complication reporting was sparse with two exceptions: (i) van Steenberghe et al. who reported 6/27 cases of bruxism (involuntary grinding of the teeth), a complication known to be associated with immediate loading of implants in fully endentulous maxillae [50], and (ii) F€ urnstahl et al. [10] who describe severe pain postoperatively due to incomplete procedure [medial compartment of the knee not repaired in theatre, despite having achieved repair during rehearsal]. The number of participants was too small (n ¼ 3) to embark on further analysis ( Table 2).

Quality assessment of included studies
The quality of the studies was appraised using the Cochrane Collaboration's Tool for Assessing Bia [61], the Newcastle -Ottawa Quality Assessment Scale (NOQAS) [62] and three minute case series appraisal tool [63] for RCTs, cohort and case control studies and case series respectively (Figure 3).

RCTs
Introduction of bias to the study conducted by Small et al. [38] is very unlikely because enough information is given regarding the randomisation and the concealment allocation processes. The assessors of the outcome were blinded and there was sufficient completeness of outcome data. The authors reported both positive and negative results, reducing the  [55] Resection margins negative for cancerous tissue No urological complications Kusaka et al. [56] NIP Li et al. [58] Surgeon's opinion: anatomy accurately represented Li et al. [51] Shorter op time and blood loss in intervention group No difference in complication rates Schepers et al. [24] 6/7 flaps survived (average f/u 9.7 months) 1.2-2mm discrepancy between pre and postoperative model Steinbacher [26] NIP Vlachopoulos et al. [40] All osteotomies healed after mean time of 3.6 months Increased ROM and grip strength for both groups Figure 2. Above: Similarity between anatomy and surgical plan during rehearsal compared with real procedure. Size of circles is indicative of sample size. Below: Face and content validity of preoperative preparation processes.
possibility of reporting bias. There do not, however, appear to have taken into account personnel blinding, which would be practically very difficult [38]. In comparison, Ayoub et al. [15] do not provide information on how randomisation took place or if there was allocation concealment or personnel, and patient and assessor blinding. In regards to the reported outcomes there were data provided about the clinical results but not about the accuracy of the pre-operative preparation technique. Finally, introduction of reporting bias is unlikely because both positive and negative results are reported [15].

Case-series and non-randomised trials
The majority of case series and non-randomised trials are well designed as shown in Figure 3. All nonrandomised trials scored six or more stars out of maximum nine and most case series had more than five (out of maximum eight) elements of welldesigned case series.

Meta-analysis
Meta-analysis was applied where permitted by similarity of the research question and the measured  outcomes. Hislop et al. [44], Desender et al. [41] and [45] all assessed the likeness of rehearsals and real procedures in vascular surgery and employed three common outcomes to do so (i.e. time to complete procedure, fluoroscopy time and fluoroscopy volume) ( Figure 4). The results of the meta-analysis demonstrate that although rehearsals were significantly quicker than real procedures (SMD ¼ À1. 56  Li et al. [51], Ayoub et al. [15], Small et al. [38], and Hsu et al. [21] recruited two distinct groups of patients to compare the clinical results of surgical procedures that were rehearsed to standard surgical treatment, while Nam et al. [9] compared the former to real time computer assisted surgery. The pre-rehearsed operations were completed in significantly less time (SMD À0.47 [À0.79, À0.16] P ¼ 0.003) but the immediate clinical outcome was similar for practiced and not practiced operations (SMD ¼0.03 [À0.23, 0.29] P ¼ 0.82) ( Figure 5).

Discussion
This review demonstrates that patient specific pre-procedural are feasible and safe. The accuracy of the anatomical models and the immediate clinical outcomes of the rehearsed procedures were satisfactory. Metaanalysis comparing rehearsals and real procedures showed good correlation, whilst when operations following rehearsals were compared to non-rehearsed ones the former were performed quicker and yielded the same surgical outcome. Practical issues in the design and manufacturing of anatomical models are also highlighted in this review and can act as focus points for future studies.
The potential of patient-specific pre-procedural rehearsals shown in our review is also highlighted in the current literature. For example, [64] conducted a review on the use of 3D printing in medicine, giving a favourable opinion regarding the usefulness of additive manufacturing techniques for patient-specific preoperative planning and for building bespoken implants and prosthetics. However, the authors recognise that their conclusions are limited by the small number of cases currently reported and should be validated by larger scale trials.
Similarly, Pratt et al. [65] looked into preparation prior to foetal surgery, suggesting that although early evidence is promising, the exact impact on surgical outcomes cannot be accurately determined due to small numbers of reported cases. A beneficial impact was also reported by See et al. [66], investigating the impact of endovascular surgery rehearsals on trainees' performance. The lack of randomised controlled trials was highlighted in both studies.
This systematic review has some limitations. Only two RCTs are included in this review whilst a significant volume of the presented evidence stems from case-series and non-randomised trials. This is a frequent occurrence in literature regarding emerging technologies in surgery [67]. Whilst some studies have shown case series to have a wider range of measured outcomes and estimates of effect size when compared to RCTs [68,69], and to report the best achievable outcome rather then the one routinely accomplished [70], others do not find these conclusions to be true. For instance, Linde et al. [71] showed non-randomised trials to have consistent results to RCTs. Similarly, Dalzier et al. found a good correlation of case series and RCT results [68]. Nevertheless, case-series can demonstrate feasibility and safety.
Application of meta-analysis was limited by the diverse outcome measures employed in various studies and was reflected in the high heterogeneity found mostly in the comparison of two distinct groups of patients (I 2 test 80% an 86%). This could be due to various reasons, one of them being small sample size in each study, which is known to be associated with high heterogeneity [72]. Further, the potential variance in population, surgical specialty and intervention implementation may confound the results [73]. However, as we are inquisitive about the application of PS rehearsals in all surgical specialties we felt it was important to continue with this analysis and report the results.
Although the introduction of pre-operative rehearsals in routine clinical care should be preceded by the provision of high level evidence, they are worth exploring because they may lead to increased efficiency and reduced operating time, whilst maintaining safe outcomes. Rapid technological advances are expected to augment the accuracy of these procedures and reduced manufacturing times for virtual and synthetic anatomical models may increase the applicability of patient-specific preprocedural rehearsals.

Funding
Leeds Teaching Hospitals Charitable Foundation