Can CAS Improve Outcomes in Knee Replacement?

    A recently published study found statistically significant improvement in some patient-reported outcome measures for computer-assisted versus conventionally instrumented total knee replacement. But is that enough reason for cautious optimism?


    William Petersen, MD, MPH, and William J. Long, MD, FRCSC


    Petursson G, Fenstad AM, Gøthesen Ø, et al. Computer-Assisted Compared with Conventional Total Knee Replacement: A Multicenter Parallel-Group Randomized Controlled Trial. J Bone Joint Surg Am. 2018 Aug 1;100(15):1265-1274. doi: 10.2106/JBJS.17.01338.


    In this article, Petursson et al present 2-year follow-up data on 167 patients who were randomly assigned to either a computer-assisted surgery (CAS) total knee replacement (TKR) or conventionally instrumented TKR. The primary outcome measured was the minimum important change difference (MICD) in the Knee Injury and Osteoarthritis Outcome Score (KOOS) 42-item, self-administered questionnaire, with the MICD being an 8- to 10-point difference for the averaged KOOS.

    This study is an extension of previously published data from the same group, in which the primary outcome reported was clinical results analyzed with radiostereometric analysis at 3 months. [1] In that study, the researchers concluded that with regard to mechanical alignment and positioning of the prosthesis, CAS TKR was more predictable, with few outliers, than conventional TKR.

    The goal of Petursson et al was to show that reproducible component alignment with CAS improves patient-reported outcome measures (PROMs). Previous large registry studies – including studies using data from the New Zealand Joint Registry and from the Norwegian Arthroplasty Registry – have failed to show such outcomes. Data from the Australian National Joint Replacement Registry showed only slightly fewer revisions in CAS TKRs at 9-year follow-up in patients over age 65.

    For their study, Petursson et al included patients who were age 50 to 85 and had inflammatory arthritis and primary or secondary osteoarthritis of the knee indicated for a TKR. They excluded patients who had:

    • American Society of Anesthesiologists (ASA) category of more than 3
    • Severe neurologic disease
    • Dementia
    • Previous cancer
    • Body mass index (BMI) above 35 kg/m2
    • Previous tibial or femoral shaft fracture
    • Severe preoperative valgus alignment (more than 15° from mechanical leg axis)
    • Previous tibial or femoral osteotomy
    • Recent knee injury (less than 1 year preoperatively)
    • Severe ipsilateral hip stiffness
    • Ipsilateral hip replacement
    • Metal allergy

    The study describes alignment and rotational position of the implants and presents 2-year data on:

    • KOOS improvements
    • Knee Society Score (KSS)
    • EuroQol-5 Dimensions (EQ-5D)
    • Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC)
    • Visual analog scale (VAS)

    The researchers also examined the Outcome Measures in Rheumatology-Osteoarthritis Research Society International (OMERACT-OASRI) criteria to calculate responder rates at 2 years using WOMAC scores.

    They report no difference between the 2 groups with respect to baseline characteristics. The CAS group had statistically and MCID significant improvements in 2 of the 5 subscales of the KOOS PROMs (symptoms and sport/recreation subscales). The group differences were:

    • 4.6 points for pain (P=0.14)
    • 7.4 points for symptoms (P=0.02)
    • 5.5 points for activities of daily living (P=0.07)
    • 16.2 points for sport and recreational function (P<0.01)
    • 6.1 points for quality of life (P=0.11)

    With regard to the WOMAC only, the stiffness subscale showed a statistical difference of 8.8 points (P = 0.03). However, there were no other significant group differences in terms of improvements in the other WOMAC subscales or the KSS and EQ-5D scores.

    Clinical Relevance

    Petrusson et al report short-term clinical outcomes of a multi-center, blinded, randomized controlled trial (RCT) comparing CAS TKR with conventionally instrumented TKR. This is the first study to demonstrate statistical and clinical MCID improvement in 2 KOOS subscales favoring CAS TKR at a 2-year follow-up. The researchers also utilized OMERACT-ARSI criteria to place WOMAC score changes into 3 categories of high responders, moderate responders, and non-responders over the 2-year period. They demonstrated that the CAS group had significantly more high responders than the conventionally instrumented group.

    In contrast with previous RCTs, this study’s early clinical results show tangible improvement in PROMs associated with CAS TKR compared with conventionally instrumented TKR. It is important to note, however, that these improvements were seen in only 3 subscales of the KOOS and WOMAC scores. All other PROMs measured were similar between groups over the 2-year period.

    In subgroup analysis, Petrusson et al compared postoperative mechanical malalignment (> 0° ± 3°) versus well-aligned knees (< 3° of varus or valgus). They found no difference in PROMs between the 2 groups at 2 years. They also looked at postoperative scores of patients who had constitutional varus before surgery (a subgroup of 82 patients) and found no significant difference in postoperative KOOS scores between those left in varus alignment and those corrected to neutral alignment. Both subgroup analyses undermine the concept that neutral anatomic alignment, which is facilitated by CAS, leads to functional gains in PROMs.

    This study has a number of shortfalls. The small number of patients and the high loss to follow-up at 2 years (more than 12%) are a concern. Randomization included a sham incision but not a drill hole at the site of CAS pins, thus allowing potential bias in the evaluations. The significant difference in stiffness (P=0.03) suggests there may be an issue with soft tissue balancing in conventional knees, which is not an issue directly related to CAS. A subgroup analysis of surgeon performance could be done to further evaluate this finding.

    Perhaps most importantly, the authors define CAS as a “method of alignment,” then compare well-aligned to poorly aligned knees, finding no difference in outcomes. Thus, the primary purpose of CAS – improved alignment – is not associated with improved outcomes. The authors point out that the numbers were small in this comparison, but as this is the very foundation for the argument, it should have been considered in their pre-study power analysis.

    The study findings demonstrate promising early functional results with the use of CAS TKR and possible support to the more widespread adoption of this method for TKR. However, for the reasons stated above, it is difficult to extrapolate that the gains in function seen in this short-term follow-up can be attributed solely to using CAS. More research is needed to better understand the ideal alignment and the use of CAS to achieve such targets that translate to better clinical outcomes.

    Author Information

    William Petersen, MD, MPH, is a fellow in adult reconstructive surgery at NYU Langone Orthopedic Hospital, New York, New York. William J. Long, MD, FRCSC, is Clinical Associate Professor of Orthopaedic Surgery, Division of Adult Reconstruction, Department of Orthopedic Surgery, at NYU Langone Orthopedic Hospital, New York, New York.


    Dr. Long has disclosed that he is a consultant with a number of companies regarding computer-aided and robotic knee surgery, including Johnson & Johnson and THINK Surgical.


    1. Gøthesen Ø, Espehaug B, Havelin LI, et al. Functional outcome and alignment in computer-assisted and conventionally operated total knee replacements: a multicentre parallel-group randomised controlled trial. Bone Joint J. 2014 May;96-B(5):609-18. doi: 10.1302/0301-620X.96B5.32516.