CORR Insights®: Are There Differences in Accuracy or Outcomes Scores Among Navigated, Robotic, Patient-specific Instruments or Standard Cutting Guides in TKA? A Network Meta-analysis

Where Are We Now?

Nearly one million TKAs are performed annually in the United States, with models suggesting this volume will approach two million by the end of the 2020s [7]. Despite interest in joint-preserving therapies such as stem cells [8], the TKA burden in the coming decades is likely to remain substantial. Osteoarthritis is increasing in frequency more quickly than can be explained by population changes in aging and obesity [11]. For all of these reasons, the expected magnitude of the demand for TKA underlies the importance of seeking small clinical improvements and the need to consider costs.

A large reason TKA has become so common is its clinical success, which comprises two sometimes overlapping but distinct components. First, there is patient satisfaction with improved quality of life, and second, there is implant survival, which reflects the avoidance of complications that result in revision. Although patient satisfaction is usually assessed within 2 years of surgery, implant survival is measured over decades. Although up to one in five patients who undergo TKA are dissatisfied in the short-term [3], implant survival remains greater than 80% at 25 years [4]. As these issues are being researched, implant manufacturers, which are often responsible for the instruments and technology used to prepare the bone, are under pressure to generate profits. In the current regulatory environment, this often results in new technology that is approved not because it improves satisfaction or implant survival, but simply because the new technology is “substantially equivalent” to some existing technology [9]. These “advances” often add cost without discernible benefit, which is a serious issue in light of increases in the prevalence of osteoarthritis and the frequency with which it is treated by TKA [8, 11].

A study reported that an abundance of small studies have examined the associations between surgical factors and short-term outcomes such as radiographic parameters [6]. Although modern analytic techniques such as network meta-analyses can maximize our use of such data, there are limits [5]. These methods can increase the precision around estimates, produce estimates for some unstudied associations, and even explore effect modifications based on study characteristics. However, even the best meta-analysis of studies on the association between bone-preparing technologies and coronal alignment cannot assess how variation in alignment is associated with unreported outcomes. A lack of primary data regarding implant survival cannot be overcome with these analytic techniques.

Acknowledging the importance of preoperative planning and having a target alignment for bony preparation is not an endorsement of alignment at any cost or an acceptance of its importance. Before applying knowledge of how technology is associated with precision in alignment, it should be acknowledged that the target coronal alignment is debated, alignment varies by the measured angle, dichotomizing alignment at a threshold of 3° of neutral is somewhat arbitrary, alignment of the femoral and tibial components are separate, and associations between these measures of alignment and implant survival are unknown [1]. Additionally, the costs and risks associated with different technologies must be considered in deciding on their appropriate use.

In this issue of Clinical Orthopaedics and Related Research®, a network meta-analysis found that computer navigation decreased the proportion of patients with alignment more than 3° from neutral, but computer navigation was not associated with a corresponding improvement in validated outcomes scores [2]. Based on this, the authors questioned whether novel approaches to achieving TKA alignment were worth the costs and added surgical time, and concluded that until these approaches are associated with better survivorship or substantially improved outcomes scores in future studies, surgeons and hospitals should not use these approaches because they add both cost and risks that are associated with the uncertainty of novel surgical approaches.

Where Do We Need to Go?

Although Bouche et al. [2] show that different bone preparation technologies are associated with precision in alignment but not patient-reported outcome scores, this is not the same as showing that alignment is not associated with patient satisfaction or implant survival. We need to understand whether alignment resulting from preparation of the femoral and tibial bone is associated with these outcomes to better inform precision of alignment. In these pursuits, the alignment measure and outcome need to be specific. Femoral, tibial, and overall alignment are not the same and may manifest differently. Whether these alignments are measured relative to an anatomic or mechanical axis matters and confuses the summarization of results across studies. The use of arbitrary thresholds of dichotomization should be discouraged. Given the number of potential associations these various exposures will create, Type I errors cannot be ignored, and research replicating any one association should be highly valued.

Researchers and readers must also consider specific outcomes, because patient satisfaction and implant survival are not the same. Although unlikely, the alignment most strongly associated with patient satisfaction could be associated with decreased implant survival. Additionally, statistical significance or the lack thereof cannot be confused when studies report on multiple outcomes. A study powered to detected differences in alignment may be underpowered to assess differences in patient-reported outcome measures if the magnitude of this association is smaller. Similarly, a study powered to detect changes in a continuous patient-reported outcome measure at 2 years is likely to be underpowered to assess differences in implant survival.

Assuming a new technology being tested is allocated to groups equal in size, powering a study to 80% to determine whether it reduces the proportion outside a range from 20% to 10% at an alpha threshold of 0.05 would require roughly 400 participants. Under the same assumptions but with the outcome changed to reducing aseptic loosening from 8% to 4%, roughly 1100 participants would be required. A study of aseptic loosening would also require additional personnel and energy to follow patients for a longer duration.

Despite these challenges, if a measure of alignment could be validated as a surrogate marker for a clinically important outcome such as aseptic loosening, which is less frequent and longer-term, then it might be possible to compare TKA techniques without having to follow larger numbers of patients for decades. The potential for effect modification may still exist, and the association between alignment and aseptic loosening or another outcome could vary based on other factors such as the fixation mode, implant design, or patient characteristics. Nevertheless, understanding the association between alignment and outcomes would be an important step in helping to improve our ability to study rapidly evolving technologies, some of which may be off the market before long-term outcomes are known.

How Do We Get There?

Unfortunately, an easily interpreted, randomized controlled trial is not the solution. It is not possible to randomize patients to a malalignment group. Although 3° or even 5° may not matter, more-severe outliers may impact how a patient perceives his or her knee as well as implant survival. Additionally, the resources required to follow these patients for many years may be impractical. Obtaining these data through a prospective cohort would require a large effort to screen patients to identify a sample with adequate variation in alignment. Such a study would be similarly limited by the difficulties of obtaining adequate follow-up data. Fortunately, there may be efficient strategies to use data that have already been collected to produce multiple sources of results that, although not perfect, could be interpreted together to enhance our understanding.

Several registries have already obtained the follow-up data needed to assess implant survival and even specific complications such as aseptic loosening. These registries, however, lack alignment data. Nested case-control studies could be performed retrospectively using patients, but these types of studies may be limited by the existence of old radiographs and would require substantial resources to locate and read them. Adding this prospectively would require resources to obtain standardized radiographs and manual measurements. This might also necessitate practice standardization that is not otherwise clinically indicated. In the near future, automating the identification of appropriate radiographs and measurement of alignment with deep-learning techniques could reduce the burden of screening radiographs and measuring alignment. Digital images would still need to be uploaded, which would pose logistical challenges including patient privacy, data storage and integration, and data ownership.

Some high-volume institutions may be able to harness their own data to inform our understanding of alignment. This would remove some but not all of the challenges associated with patient privacy and data sharing. Similar to registries, manually recording alignment data would still be resource-intensive. Many institutions may lack the long-term outcome data that registries obtain, so researchers should consider methods of validating longer-term follow-up studies. Individual researchers should also consider how their results may be reanalyzed in the future. The best practice in this respect would be to publish the original data along with the article, although this has not yet become routine [10]. What we do in the absence of a true understanding of how alignment is associated with patient satisfaction and implant survival is likely to fuel continued debate on the use of new technologies for bone preparation in TKA. The results presented by Bouche et al. [2] clarify that currently available data do not justify the use of new bone preparation technologies such as navigation, based on improved patient-reported outcomes.