Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

Patient‐specific cutting guides for total knee arthroplasty

Authors' declarations of interest

Version published: 13 March 2017 Version history

https://doi.org/10.1002/14651858.CD012589
Collapse all Expand all

Abstract

This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:

To assess the potential benefits and possible harms of patient‐specific cutting guides versus conventional instruments or computer‐assisted surgery for total knee arthroplasty. We will examine improvements in lower limb alignment and other clinical outcomes in participants undergoing TKA surgery.

We will conduct this review according to the guidelines recommended by the Cochrane Musculoskeletal Group Editorial Board (Ghogomu 2014).

Background

Description of the condition

Osteoarthritis, also known as degenerative arthritis, is a group of diseases and mechanical abnormalities involving degradation of joints. Osteoarthritis is the most common disease of joints in adults worldwide, and the knee joint is the most common major joint to be affected (Heijink 2012). The degenerative process predominantly affects the articular cartilage, which is a smooth, white tissue that covers the ends of bones where they form joints. A variety of causes including hereditary, metabolic, developmental and mechanical conditions can initiate the process of cartilage degeneration within the joint. Ageing contributes to, but does not directly cause, osteoarthritis. Obesity, female sex and a low level of education are other significant risk factors for the development of osteoarthritis (Andrianakos 2006).

Individuals with osteoarthritis usually present with pain and loss of motion. The pain is usually a dull ache that occurs during weightbearing activities such as walking, and it gradually progresses over the years. As the articular cartilage degenerates, the underlying bone is exposed, and the bone surfaces become uneven. The cartilage debris also initiates inflammation within the joint. In advanced stages, the underlying bone becomes thick and deformed, producing bony outgrowths called osteophytes. Advanced osteoarthritis can result in limb deformity, joint contractures and muscle atrophy (O'Reilly 1998; Zhang 2010).

Rheumatoid arthritis, on the other hand, is an autoimmune, inflammatory disorder that can affect multiple tissues and organs but predominantly affects synovial joints. The body's immune system, which protects human health by attacking foreign particles such as bacteria and viruses, mistakenly attacks the body's own tissues (Firestein 2003). Usually the disease affects the small joints of hands and feet. However, knee joint involvement is also common, and if left untreated, it can lead to severe disability. Other uncommon causes of inflammatory arthritis of the knee include ankylosing spondylitis, Reiter's disease and seronegative spondyloarthropathy.

Post‐traumatic arthritis can occur following trauma to the joint surfaces of the bones forming the knee joint. Trauma that results in an irregular joint surface can lead to abnormal loading of the joint, cartilage degeneration and finally osteoarthritis.

Description of the intervention

Total knee arthroplasty (TKA), also known as total knee replacement, is a surgical procedure by which the surfaces of the knee joint are replaced with metal and plastic implants (Singh 2013). It is the procedure of choice when more than one compartment of the knee is irreversibly damaged due to arthritis and symptoms are severe. The number of primary TKAs performed in the United States doubled in the last decade, with more than 97% performed for osteoarthritis (Losina 2012; Weinstein 2013). Knee replacement is successful in relieving pain and improving function in individuals with advanced arthritis (Carr 2012). However, complications such as infection, instability, loosening, dislocation or fracture can occur in around five per cent of patients in the initial 6 months following this procedure (Kane 2003). In addition, improper alignment of prosthetic components increases the risk of failure in due course and subsequent revision (Fang 2009).

In total knee arthroplasty surgery, several cuts (resections) are made in the lower end of the thigh bone (femur) and upper end of the shin bone (tibia), which together form the knee joint, to allow fitting of the femoral and tibial prostheses. Accurately aligning these cuts determines alignment of the prosthesis in the sagittal, coronal and axial (rotation) planes. A number of techniques are available to assist in aligning the components in total knee arthroplasty, including conventional techniques (standard cutting guides), patient‐specific cutting guides and computer‐assisted navigation (MacDessi 2014).

In the conventional techniques, bony resections and hence the alignment of components is determined by using visual anatomic landmarks and the anatomic axes of femur and tibia to place the standard (conventional) cutting guide. These techniques are prone to error, as they are not based on an accurate assessment of the overall alignment of the mechanical axes of the lower limb (Iorio 2013; Laskin 2003) In addition, conventional techniques are difficult to use in knees with extra‐articular deformities (deformity above or below the knee joint) (Catani 2012; Lee 2014; Tigani 2012).

The computer‐assisted navigation technique involves the creation of a digital image that guides the surgeon in component positioning during surgery (Desai 2011). Various tools are available to create the digital image such as preoperative imaging, intraoperative radiographs and image‐free techniques. The advantages of computer‐assisted techniques include dynamic assessment of deformity during surgery, calculation of soft tissue tension and accurate component positioning. However, disadvantages such as prolonged operative time, high cost and a steep learning curve have prevented this strategy from gaining widespread acceptance (Bauwens 2007).

Patient‐specific cutting guides (PSG) or custom‐made cutting guides are a new technology used in total knee arthroplasty whereby preoperative imaging (magnetic resonance imaging (MRI), computerised tomography (CT) scans and radiographs) are used to create anatomic 3D models of the patient's knee (Ast 2012). PSGs are manufactured based on these models and applied to the knee joint during surgery to determine the resection angles made to the femur and tibia. Using PSGs obviates the need to place pins in the shaft of the femur and tibia (often used in computer navigation) and the need to open the intramedullary canal of bone (used in the conventional technique).

How the intervention might work

People requiring TKA often have some degree of deformity that contributes to their pain and poor function. Correction of these deformities through restoration of the the neutral mechanical axis of the lower limb during TKA has proven to be a factor in improving functional outcomes (Choong 2009; Huang 2015). Even with experienced surgeons, coronal malalignment using conventional cutting guides can occur in approximately 28% of participants (Ng 2012). The major postulated benefit of using either PSGs or computer‐assisted navigation in TKA is that it will improve the accuracy in restoring normal knee alignment compared to conventional techniques (Bäthis 2004; Ng 2012). Better alignment of components has been correlated with improved functional outcome and long‐term success of total knee arthroplasty (Jeffery 1991). The goal of TKA is to restore neutral alignment of the mechanical axes of the limb. Replaced knees with varus alignment (bow leg deformity) fail primarily by medial tibial collapse whereas valgus knees (knock knee deformity) fail due to ligament instability (Fang 2009). Wh ile PSGs help in proper component positioning, they are not a substitute for other predict ors of successful TKA such as car eful preoperative planning, good intraoperative judgement , appropriate soft tissue balancing and precise implantation techniques ( Ng 2012).

Patient‐specific cutting guides may also reduce the duration of surgery and incision length (Chareancholvanich 2013; Noble 2012; Nunley 2012a). Avoiding penetration of the intramedullary canal while using PSGs or computer‐assisted navigation may reduce bleeding (Kalairajah 2005). Performing bilateral simultaneous TKA is associated with high incidence of mortality and pulmonary embolism compared to unilateral or staged procedures, and one plausible explanation is the higher embolic load of fat and marrow elements generated while penetrating the intramedullary canal (Parvizi 2001; Restrepo 2007). However, whether these advantages have any clinical relevance needs to be studied.

Potential disadvantages of patient‐specific cutting guides including the high cost of manufacture, the need for preoperative MRI or CT and the influence of a learning curve (Nunley 2012a; Victor 2014). Conversely, computer‐assisted navigation might increase the surgical time (Bäthis 2004). Also, some studies have concluded that PSGs and computer‐assisted navigation do not offer any additional benefit compared to conventional cutting guides in restoring the normal knee alignment (Chareancholvanich 2013; Kim 2007; Marimuthu 2014; Parratte 2013; Victor 2014). Several authors have reported a need to switch to conventional technique intraoperatively when PSG‐based resections deviated excessively from the resections based on conventional or navigation guides (Roh 2013).

Why it is important to do this review

Despite inconsistent evidence validating its effectiveness, there are increasing reports in the literature on the use of patient‐specific cutting guides for total knee arthroplasty (Noble 2012; Victor 2014). Some investigators have postulated that PSGs improve the alignment of components by reducing variance (Ng 2012). Studies comparing these guides to conventional instrumented techniques differ widely in their findings (MacDessi 2014; Marimuthu 2014; Noble 2012; Victor 2014).

In a retrospective comparative analysis, Ng 2012 found that the normal mechanical axis of the limb was restored more often in the PSG group than the conventional method and there were fewer outliers in the PSG group. The mechanical alignment of the limb measured by hip–knee–ankle angle is the angle between the mechanical axis of the femur and tibia. The mean difference in hip–knee–ankle angle was not statistically significant between the two groups. Considering individual component alignment in the coronal plane, the angles were closer to neutral in the PSG group (femur P = 0.005, tibia P < 0.001), but the proportion of outliers was similar in both groups. In a retrospective analysis, Daniilidis 2013 found that the PSG group had fewer outliers in coronal HKA than in the conventional group, and in a different retrospective analysis of femoral component rotation based on MRI, Heyse 2012a found that PSG resulted in fewer outliers than the conventional group.

Noble 2012 reported preliminary results of a randomised controlled trial (RCT) comparing PSG with conventional techniques and found better coronal alignment (closer to neutral) in the PSG group. Further, they found that the PSG group had a shorter operative time, length of hospital stay and incision length and used fewer instrument trays. Other RCTs found no significant differences between the conventional and PSG groups in terms of mean HKA and femoral or tibial coronal alignment (Boonen 2012; Chareancholvanich 2013; Parratte 2013; Roh 2013).

When analysing outliers in RCTs, Chareancholvanich 2013 reported a reduction in the number of coronal femoral alignment outliers with PSG, whilst Victor 2014 reported a reduction in coronal tibial outliers with conventional TKA. RCTs by Roh 2013, Boonen 2013 and Parratte 2013 did not find any difference in the number of coronal alignment outliers between the two groups.

Boonen 2013, Parratte 2013, Roh 2013 and Victor 2014 evaluated sagittal femoral and tibial alignment in RCTs. There was no difference between the two groups in terms of mean sagittal femoral or tibial component alignment in these studies. Boonen 2013 reported a significantly lower number of sagittal femoral outliers with conventional method, and Victor 2014 reported fewer outliers for tibial component sagittal alignment. Roh 2013 and Parratte 2013 did not find any significant difference between the two groups in terms of sagittal outliers.

Parratte 2013 did not find any significant difference between the two groups in terms of femoral and tibial component rotation both in terms of mean angle and percentage of outliers.

While there are multiple studies reporting a lack of advantage for computer‐assisted surgery over conventional techniques (Allen 2014; Kim 2012), we could not find any studies in the literature comparing PSG with computer‐assisted surgery in total knee arthroplasty.

Thus, we are conducting this systematic review to explain the uncertainty due to conflicting results in the studies. The aim of our systematic review is to determine whether the use of PSGs in total knee arthroplasty offers an advantage over conventional methods.

Objectives

To assess the potential benefits and possible harms of patient‐specific cutting guides versus conventional instruments or computer‐assisted surgery for total knee arthroplasty. We will examine improvements in lower limb alignment and other clinical outcomes in participants undergoing TKA surgery.

We will conduct this review according to the guidelines recommended by the Cochrane Musculoskeletal Group Editorial Board (Ghogomu 2014).

Methods

Criteria for considering studies for this review

Types of studies

In this review, we will include RCTs comparing conventional instrumentation or computer‐assisted surgery versus patient‐specific cutting guides for TKA. We will include studies reported as full‐text, those published as abstract only, and unpublished data, and there will be no restrictions on length of follow‐up or language of the paper. Where full text reports are not available, we will request the authors for their results or leave the study in 'ongoing studies' if full results are awaite d.

Types of participants

Adults undergoing primary total knee arthroplasty for advanced arthritis due to any cause, including osteoarthritis, post‐traumatic arthritis, rheumatoid arthritis or other inflammatory arthritis will be eligible for inclusion. We will exclude participants receiving knee replacement for tumours, unicompartmental replacement or revision knee replacement. We will also exclude participants undergoing total knee arthroplasty following either previous osteotomy or previous unicompartmental replacement arthroplasty on the same knee.

Types of interventions

We will include trials comparing total knee arthroplasty performed using patient‐specific guides versus conventional instrumentation or computer‐assisted surgery. Conventional instrumentation involves use of intramedullary and/or extramedullary alignment guides as well as anatomical landmarks to help position the cutting guides (Victor 2014). This system can be used in any patient, unlike patient‐specific guides (PSG), which are individually designed based on the patient's anatomy for use with a particular knee prosthesis. We will consider different implant designs (posterior stabilised, cruciate retaining and cruciate sacrificing) and different bearing mechanisms (fixed bearing and mobile bearing). We will also include all modes of fixation of prostheses, including cemented, non‐cemented and hybrid fixation (combination of cemented and non‐cemented techniques for different parts of the replacement). Studies will also be included irrespective of whether patellar resurfacing is performed.

Types of outcome measures

Major outcomes

  1. Survival rate of the implant (any revision surgery to change a prosthetic component, i.e. the proportion of knees that undergo revision surgery)

  2. Functional outcome measures with validated instruments (such as participant reported measures: Western Ontario & McMaster Universities Osteoarthritis Index (WOMAC), Knee Injury and Osteoarthritis Outcome Score (KOOS), International Knee Documentation Committee (IKDC) score, Oxford Knee score; and clinician reported: Knee Society Score)

  3. Radiological alignment of component: we will accept the femorotibial coronal angle (FTCA) and other radiographic data points measured following surgery (any time within three months). If investigators perform the measurements at more than one time point, we will use the earliest postoperative measurement; we will measure the outcome as a continuous variable in degrees

  4. Pain: mean overall pain measured on a visual analogue scale (VAS) or numerical rating scale (NRS), or other measures of pain

  5. Global assessment (participant‐reported) (e.g. better, not better, worse, dichotomised to better and not better or worse)

  6. Total adverse events (including infection, thrombosis, palsy, death, etc.)

  7. Re‐operation rate (not involving implant change)

Minor outcomes

  1. Quality of life

  2. Operative time

  3. Blood loss

  4. Range of motion

  5. Length of hospital stay

  6. Incision length

  7. Number of outliers in radiographic alignment (in the coronal plane); outlier is defined as deviation of more than three degrees from the target (Victor 2014); we will measure this as a dichotomous variable

Search methods for identification of studies

Electronic searches

We will search the following databases, with no language, time or publication type restrictions.

  • Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library) using the search strategy outlined in Appendix 1.

  • MEDLINE (Ovid) using the search strategy outlined in Appendix 2.

  • EMBASE (Ovid) using the search strategy outlined in Appendix 3.

We will search the World Health Organization International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/en) and ClinicalTrials.gov (clinicaltrials.gov) for ongoing and unpublished trials. If we identify any ongoing trials, we will contact authors and request full trial reports for inclusion in the review.

Searching other resources

We will handsearch the reference lists of retrieved studies and those of narrative and systematic reviews to find additional potentially relevant studies. We will also contact the authors of included studies for additional information, as necessary.

Data collection and analysis

Selection of studies

Two review authors (VSK and KM) will screen all the titles, abstracts and keywords of publications identified by the search to assess their eligibility. We will exclude publications that clearly do not meet the inclusion criteria at this stage. We will retrieve the full text of all potentially relevant papers, and two review authors (VSK and KM) will independently review complete copies of each study and indicate on a study eligibility form if the study should be included, excluded or if they are undecided. We will resolve disagreement regarding study inclusion by discussion between the two review authors, and, if necessary, by the involvement of a third reviewer (SM). We will contact the corresponding author for clarification if it is unclear whether a trial is eligible for inclusion.

Data extraction and management

We will extract and record data using data extraction forms, which will be developed and piloted by two authors (VSK and KM). We will resolve any differences through discussion with the senior authors (SM and DC). We will attempt to obtain any missing data from the corresponding authors. We will compile the following information from the included studies.

  • General study information (e.g. title, authors, publication year, country).

  • Characteristics of the study: design, study setting, inclusion and exclusion criteria, risk of bias (using Cochrane 'Risk of bias' tool).

  • Characteristics of the study population and baseline characteristics of the intervention (PSG) and control groups (conventional instrumentation or computer‐assisted surgery) (age, sex, duration of disease, concurrent treatments) and numbers in each group.

  • Type of prosthesis used (cruciate retaining or cruciate sacrificing).

  • Type of bearing surface (fixed bearing or mobile bearing).

  • Outcome measures used.

  • Withdrawals.

  • Loss to follow‐up.

  • Length of follow‐up.

We will use this information to populate the 'Characteristics of included studies' table for each included study.

If trials report more than one measure for pain, we will use a hierarchy of seven levels, using the highest measure on the following list (e.g. we will choose overall pain over pain on walking).

  1. Overall pain (global pain), on a visual analogue scale (VAS) score, or numerical rating scale (NRS).

  2. Pain on walking.

  3. WOMAC osteoarthritis index pain subscore.

  4. Pain on activities other than walking.

  5. Rest pain.

  6. Other algofunctional scale.

If trials report more than one measure for physical function (participant reported), we will choose a continuous scale over an ordinal scale and use one of the following scores, selecting the measure highest on the list if two or more are reported in a single study, but we will also report all scores used.

  1. Knee injury and Osteoarthritis Outcome Score (KOOS score).

  2. Oxford knee score.

  3. WOMAC disability subscore.

  4. Composite disability scores other than WOMAC.

  5. Disability other than walking.

  6. WOMAC global scale.

  7. Lequesne osteoarthritis index global score.

  8. Other algofunctional scale.

For quality of life, we will accept the following scores, taking the highest on the list if two or more are reported in a single study.

  1. EuroQoL‐5 Dimensions (EQ‐5D).

  2. 36‐Item Short Form Survey (SF‐36).

  3. 12‐Item Short Form Survey (SF‐12).

  4. Other scale system.

If trialists report both final values and change from baseline values or the same outcome, we will extract change from baseline values, provided the standard deviation (SD) for the change score is also reported. If both unadjusted and adjusted values for the same outcome are reported, we plan to extract the adjusted values.

If data are analysed based on an intention‐to‐treat (ITT) principle and per‐protocol (as‐treated), we plan to extract ITT data for both benefits and harms.

If there are multiple time points, we will abstract the data for three timeframes.

  1. Short term: 24 months or less after operation.

  2. Intermediate term: 25 months to 120 months after operation.

  3. Long term: more than 120 months after operation.

If there are multiple time points within the above mentioned timeframes, we would include the data from the latest time point within the frame. For example, if a particular study reports outcomes at 12, 24 and 36 months of follow‐up, we will consider the data from 24 months as short term and that at 36 months as intermediate term, and we will not extract the 12‐month data. We will note in the Characteristics of included studies tables, which time times we have excluded from the review.

Assessment of risk of bias in included studies

Two authors (VSK and KM) will independently assess the risk of bias of each trial according to Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b), record the information in a table and provide a narrative description in the text. We will resolve disagreements by discussion or by involving a third author (IAH), and we will contact the corresponding study authors for clarification if information is unclear.

1. Sequence generation

Is the allocation sequence adequately generated?

To check for possible selection bias, for each included study we will categorise the method used to generate the allocation sequence as being at:

  • low risk of bias (any truly random process, e.g. random number table, computer random number generator);

  • high risk of bias (any non‐random process, e.g. odd or even dates of birth, hospital or clinic record numbers);

  • unclear risk of bias (method of sequence generation not reported).

2. Allocation concealment

Is allocation adequately concealed?

To check for possible selection bias, for each included study we will categorise the method used to conceal the allocation sequence as being at:

  • low risk of bias (e.g. telephone or central randomisation, consecutively numbered sealed opaque envelopes);

  • high risk of bias (open random allocation, unsealed or non‐opaque envelopes, alternation, date of birth);

  • unclear risk of bias (not reported).

3 . Blinding of participants and personnel

Are participants and personnel blinded to allocation?

To detect the presence of performance bias, we will assess whether included studies have reported adequate blinding of participants and personnel.

4. Blinding of outcome assessment

Are outcome assessors blinded to allocation?

To check for detection bias, for each trial we will consider blinding separately for subjective self reported outcomes (such as pain, function, global assessment) and objective outcomes (such as re‐operation rate) because for unblinded outcome assessment, risk of bias for objective outcomes may be different than for a participant‐reported outcomes.

5. Incomplete outcome data

Are incomplete outcome data adequately addressed?

To check for possible attrition bias through withdrawals, dropouts and protocol deviations, for each included study and for each outcome we will describe the completeness of data including attrition and exclusions from the analysis. We will note whether trials have reported attrition and exclusions, the numbers included in the analysis at each stage (compared with the total number of randomised participants), reasons for attrition or exclusion where reported, and whether missing data are balanced across groups or were related to outcomes. Where trial authors report or supply sufficient information, we will re‐include missing data in the analyses. We will try to gather the missing data by contacting authors of the trial.

If efforts to secure the missing data fail, then the missing data will be subdivided as either 'missing at random' or 'not missing at random'. We will classify data as 'missing at random' if their absence is unrelated to their actual values. We will explicitly explain the decision and its justification in the review. We will analyse the available data only. On the other hand, if the fact that a particular datum is missing could be related to its actual value, we will classify it as 'not missing at random' and explicitly explain the reasons for such classification. In such cases, we will use statistical models to allow for missing data, making assumptions about their relationships with the available data. We will also perform sensitivity analyses to assess how sensitive the results are to reasonable changes in the assumptions made, and we will address the potential impact of the missing data on the findings of the review in the Discussion.

6. Selective reporting bias

Are reports of the study free of suggestion of selective outcome reporting?

For each included study, we will describe how we investigated the possibility of selective outcome reporting bias and what we found. We will assess the methods as being at:

  • low risk of bias (where it is clear from the protocol (if available) or the Methods that investigators have reported all of the study's pre‐specified outcomes and all expected outcomes of interest);

  • high risk of bias (where investigators have not reported all the study's pre‐specified outcomes; they did not pre‐specify one or more reported primary outcomes; they reported outcomes of interest incompletely, making it impossible to use them; study fails to include results of a key outcome that would have been expected to have been reported);

  • unclear risk of bias.

Other sources of bias

Is the study apparently free of other problems that could put it at a high risk of bias?

For each included study, we will describe any important concerns we have about other possible sources of bias (for example, whether there was a potential source of bias related to the specific study design or whether the trial was stopped early due to some data‐dependent process). We will assess studies as being at:

  • low risk of other bias;

  • high risk of other bias;

  • unclear risk of other bias.

Overall risk of bias

We will make explicit judgements about whether studies are at overall high risk of bias:

  • overall low risk of bias (low risk of bias for all key domains within the study);

  • overall high risk of bias (high risk of bias for one or more key domains within the study);

  • overall unclear risk of bias (unclear risk of bias for one or more key domains within the study).

With reference to the domains 1 to 6 above, we will assess the likely magnitude and direction of the bias and whether we consider it is likely to impact on the findings. We will also consider the impact of missing data by key outcomes.

Measures of treatment effect

We will calculate overall effects from the studies for which data are available. For dichotomous outcomes (number of outliers in radiological alignment of component, total adverse events, survival of implant, re‐operation rate not involving implant exchange) we will express the result as a risk ratio (RR) with 95% confidence intervals (CI). For continuous scales of measurement (radiological alignment of component, functional measure, pain, global assessment, quality of life, operative time, blood loss, range of motion, length of hospital stay and incision length), we will present the mean difference (MD) and 95% CI, or the standardised mean difference (SMD) and 95% CI if different scales are used to measure the same outcome. When different scales are used to measure the same conceptual outcome (e.g. pain), we will back‐translate the SMD to a typical scale (e.g. 0 to 10 for pain) by multiplying the SMD by a typical among‐person standard deviation (e.g. the standard deviation of the control group at baseline from the most representative trial) (Schünemann 2011b).

In the 'Effects of interventions' Results section and the 'Comments' column of the 'Summary of findings' table, we will provide the absolute percentage difference, the relative percentage change from baseline, and the number needed to treat for an additional beneficial outcome (NNTB), or the number needed to treat for an additional harmful outcome (NNTH). We will only provide the NNT or NNTH when the outcome shows a statistically significant difference.

For dichotomous outcomes, we will calculate NNTB or NNTH from the control group event rate and the relative risk using the Visual Rx NNT calculator (Cates 2008). We will calculate NNTB or NNTH for continuous measures using the Wells calculator (available at the CMSG Editorial office; musculoskeletal.cochrane.org). We will determine the minimal clinically important difference (MCID) for each outcome for input into the calculator. We will use 1.5 points on 0 to 10 point VAS pain scale and 10 points on 100 point function scale.

For dichotomous outcomes, we will calculate the absolute risk difference using the risk difference statistic in Review Manager (RevMan) software (RevMan 2014), expressing the result as a percentage. For continuous outcomes, we will calculate the absolute benefit as the improvement in the intervention group minus the improvement in the control group, in the original units, expressing the result as a percentage.

We will calculate the relative percentage change for dichotomous data as the risk ratio minus one and express the result as a percentage. For continuous outcomes, we will calculate the relative difference in the change from baseline as the absolute benefit divided by the baseline mean of the control group and express the result as a percentage.

Unit of analysis issues

The unit of analysis will be the participant. If studies report results for bilateral surgery but have randomised by person in their analyses, without adjustment for the non‐independence between limbs, there may be potential unit of analysis errors. We will attempt to re‐analyse such studies by calculating sample sizes where possible, as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a).

Dealing with missing data

For included studies, we will note levels of attrition. When we discover missing data during data extraction, we will attempt to contact the study authors to request the required information. Where possible, we will compute missing standard deviations from other statistics such as standard errors, confidence intervals or P values, according to the methods recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011c). If we cannot calculate standard deviations, we will impute them, for example from other studies in the meta‐analysis (Higgins 2011a).

Assessment of heterogeneity

We will assess included trials for clinical homogeneity in terms of participants, interventions and comparators. We will examine heterogeneity across studies by inspecting the distribution of point estimates for the effect measure and the overlap in their confidence interval on the forest plot. We will use the I2 statistic to check the statistical consistency, defined as the ratio of between‐study variation compared to the overall variation (Deeks 2011). A value greater than 50% may be considered to be substantial heterogeneity, and if we obtain this value, we will explore the sources of heterogeneity further.

Assessment of reporting biases

In order to determine whether reporting bias is present, we will determine whether the protocol for the trial was published before recruitment of participants to the study began. For studies published after 1 July 2005, we will screen the WHO International Clinical Trials Registry Platform (ICTRP) (De Angelis 2004). We will use this information to evaluate whether selective reporting of outcomes is present (outcome reporting bias).

If there are 10 or more studies in the meta‐analysis, we will analyse reporting bias using the funnel plot method. We will assess funnel plot asymmetry visually and use formal tests for funnel plot asymmetry. If asymmetry is detected in any of these tests or is suggested by a visual assessment, we will use the 'trim and fill' method to investigate and correct the bias (Sterne 2011).

We will compare the fixed‐effect model against the random‐effects model to assess the possible presence of small sample bias in the published literature (i.e. in which the intervention effect is more beneficial in smaller studies). In the presence of small sample bias, the random‐effects model shows the intervention to be more beneficial than the fixed‐effect model estimate (Sterne 2011).

Data synthesis

We will analyse the results of the studies using RevMan 2014. If we consider studies to be sufficiently clinically homogenous, we will pool data in a meta‐analysis using a random‐effects model, irrespective of the I2 values, and perform a sensitivity analysis with the fixed‐effects model.

GRADE and 'Summary of findings' tables

We will present the major outcomes of the review in a 'Summary of findings' table: survival rate of the implant (any revision surgery to change a prosthetic component), function, radiological alignment of component (overall coronal alignment), pain, participant‐reported global assessment, total adverse events (including infection, thrombosis, palsy, death, etc.) and re‐operation rate (not involving implant change). We will present two different 'Summary of findings' tables: one for PSG versus conventional instrumentation and the other for PSG versus computer‐assisted surgery. These will provide key information concerning the quality of evidence, the magnitude of effect of the interventions examined and the sum of available data on the major outcomes, as recommended by Cochrane (Schünemann 2011a). This table will include an overall grading of the evidence related to each of the main outcomes using the GRADE approach (Schünemann 2011b).

Two review authors will independently assess the quality of the body of evidence for studies contributing data to each of the seven major outcomes. We will use the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of evidence, and report the quality as high, moderate, low, or very low. We will use methods and recommendations described in section 8.5 and 8.7, and chapters 11 and 12, of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b; Schünemann 2011a; Schünemann 2011b). We will use GRADEpro software to prepare the 'Summary of findings' tables (GRADEpro 2015). We will justify all decisions to down‐grade the quality of studies using footnotes, and we will make comments to aid the reader's understanding of the review where necessary. We will provide the number needed to treat for an additional beneficial outcome (NNTB) or the number needed to treat for an additional harmful outcome (NNTH), absolute and relative percentage change in the 'Comments' column of the 'Summary of findings' table as described above in Measures of treatment effect.

Subgroup analysis and investigation of heterogeneity

We will use subgroup analysis to explore possible sources of heterogeneity if sufficient data are available on two of the major outcomes: 'radiological alignment of component' and 'functional measures with validated instruments (WOMAC, KOOS, OXFORD, IKDC, etc.)'.

We plan the following subgroup analyses.

  1. Participants' body mass index (BMI) (BMI ≤ 25 and BMI > 25).

  2. Participants' preoperative coronal plane deformity (mild: −6⁰ to +6⁰; severe: < −6⁰ or more than +6⁰).

  3. Type of prosthesis (cruciate retaining versus posterior stabilised).

  4. Resurfacing (patellar resurfacing versus non‐resurfacing).

We will contact the corresponding author to obtain information as needed.

We will assess differences between subgroups by inspection of the subgroups' confidence intervals. Non‐overlapping confidence intervals indicate a statistically significant difference in treatment effect between the subgroups. We will also use the formal test for subgroup interactions in RevMan 2014.

Sensitivity analysis

If sufficient data are available, we plan sensitivity analyses to assess the impact of any selection bias attributable to inclusion of trials with inadequate or unclear treatment allocation concealment. Depending on availability of data, we are also planning a second sensitivity analysis to assess the robustness of the outcomes to detection bias (by restricting to studies with adequate blinding of outcome assessors).

We will perform sensitivity analyses using radiographic alignment of component and functional outcome measure.