1 Introduction

Total knee arthroplasty (TKA) can provide significant pain relief and improved function for patients with symptomatic end-stage knee osteoarthritis. However, consistent and reproducible component placement remains a particular challenge to the operating surgeon. Robotic assisted surgery is a relatively new technology which may provide certain benefits such as more accurate placement of arthroplasty components, less soft tissue damage, improved pain control, and higher early patient reported outcomes and patient satisfaction.^1–12

Some surgeons remain hesitant to adopt robotic technology, citing the time investment required to achieve proficiency and the lack of unclear benefit to long term clinical or radiologic outcomes when compared with conventional jig-assisted arthroplasty.^6,13,14 A recent study by Kim et al. found no significant differences in component radiographic parameters, implant survivorship or clinical outcomes when comparing conventional TKA with robotic assisted TKA at an average of 13 years postoperatively.¹³ Jeon et al. also retrospectively compared conventional TKA to robotic assisted TKA at an average of 10 years, demonstrating no significant difference in postoperative complications or clinical and radiographic outcomes.¹⁴ Despite this controversy, robotic assisted TKA can provide more reproducible component placement and may even promote efficiency within the surgical team once expertise is achieved.

2 Material and methods

3 Results

3.1

3.1 Procedure duration

For the primary outcome of procedure duration (in minutes), a time series analyses uses a 50-procedure moving average for which the graphical plot is shown below (Fig. 1). For this particular procedure and physician, the analyses demonstrated substantially improved operative times by the 50th completed robotic-assisted surgery, with the surgeon reaching a stable plateau in operative times after approximately two years of surgical experience, by the time this senior surgeon had completed between 151 and 200 total procedures.

Close

Fig. 1 Moving average plot for procedure duration.

Export to PPT

To further assess stages of surgical efficiency, we looked at distinct 50-procedure windows (e.g., 1–50, 51–100, 100–150, etc.) and estimated mean procedure times for these windows using ANOVA. As the time-series analysis demonstrated with clarity that plateau occurs no later than the 200th, only the first eight windows (400 patients) were used in the follow-up. Results are provided in Table 2, which indicates that the average surgical duration for the first 50 procedures performed is around 85 min, dropping to 69 min for the next 50 procedures performed, 66 min for the next 50 procedures performed, and then plateauing at about 60–62 min following 150 procedures performed. This is further supported by Tukey multiple comparisons associated to the ANOVA which indicate that Period 1 (the first 50 procedures) mean operating times are substantially higher compared to all other periods (all p-values <0.001), and that mean operating times in Periods 2 is likewise significantly higher when compared to periods 4 through 8 (all p-values <0.05). While there is no evidence that period 2 differs from period 3 (p-value 0.842), there is further evidence that period 4 has a lower average in comparison to period 3 (p-value 0.044). None of periods 4 through 8 indicated evidence of differences in mean operating times when compared with each other (all p-values greater than 0.05). Confidence interval estimates for each period are shown in Table 2.

Table 2 Procedure windows in groups of 50.

Period	Procedures	Average (SD) Procedure Time (minutes)	95% Confidence Interval
1	1–50	84.9 (12.2)	82.4 to 87.6
2	51–100	68.9 (9.9)	66.3 to 71.4
3	101–150	66.3 (9.0)	63.7 to 68.8
4	151–200	61.1 (7.7)	58.5 to 63.6
5	201–250	62.5 (10.2)	59.9 to 65.0
6	251–300	60.6 (8.4)	58.0 to 63.1
7	301–350	61.5 (7.8)	59.0 to 64.1
8	351–400	62.0 (7.0)	59.5 to 64.5

3.2

3.2 Length of stay

There was a notable trend towards increased length of stay during the first group of 50 procedures performed, but this was not found to be statistically significant (Fig. 2). The related ANOVA likewise found no evidence of any pattern of variation of average length of stay across the first eight 50-procedure periods (most p-values >0.05).

Close

Fig. 2 Moving average plot for length of stay.

Export to PPT

4 Discussion

This study goes well beyond the initial learning curve of robotic-assisted TKA to quantify procedure volume which appears to correlate with a steady state of maximum efficiency in operating times. To the author's knowledge, with 682 consecutive cases by a single surgeon over a 4-year period, this is the largest cohort of robotic assisted TKAs analyzed in an attempt to define the ceiling of robotic-assisted surgical efficiency. The results from this study highlight a significant improvement in operative duration over the first 50 procedures with subsequent stabilization in operative duration reached between 151 and 200 total procedures. There was no evidence of difference in length of stay and post-operative readmission/reoperations as the surgeon performed more cases throughout this large-scale period.

Other studies reporting on smaller series of patients had previously suggested that a learning curve may be achieved in as little as 7 to 43 cases. These studies were inherently limited by their smaller cohort size and noncontiguous time points analyzed. Kayani et al. found significantly increased operative times and high levels of anxiety in the surgical team for the first 7 cases performed using robotic assisted technology, with average procedure duration decreasing from 83.1 min (cases 1–10) to a range of 65.3–67.2 min (cases 11–60).⁵ Vermue et al. also showed a learning curve of between 11 and 43 cases depending on the surgeon volume.¹⁶ Although these studies identify a potential learning curve, our goal was to define a steady state at which proficiency in procedure duration is achieved over time.

A study by Marchband et al. reported that operative duration during robotic assisted TKA were similar to conventional TKA just 6 months into the learning curve. Even further improvement in operative duration was seen after 1 year (62 min, t = 12 months versus 81 min, t = 0 months), with the surgeon performing 88% of procedures within 50–69 min.¹⁷ However, only specific cohorts were analyzed at each time point of 0, 6, and 12 months. This study did not track every procedure continuously over the year to assess trends or a nadir in operative duration, nor did they specify the total number of procedures performed over that year. Our study not only demonstrates a stable plateau in operative duration, but also suggests that it may take up to two years for the surgeon to reach this stage, at a rate of approximately 100 robotic surgeries per year. Similar to other studies, no increased risk in perioperative complications was seen in our cohort.^5,16,18 Our data suggests that operative efficiency continues to improve over time until reaching a stable plateau, and 20 cases do not represent the full learning curve of efficiency for this technology. Indeed, the inflection point for true proficiency may be closer to more than 150 and 200 robotic-assisted surgical procedures.

The decrease in total operative duration which is seen over time may be multifactorial. Familiarity with graphical displays, registration of landmarks, and bony resection technique may contribute to decreased operative time. Furthermore, the surgical team becomes more efficient in multitasking with increasing cumulative operative experience while also troubleshooting any technical issues which may be encountered. There have been reports in which robotic assisted procedure duration may become faster and more efficient than conventional manual TKA in some hands.^17,19 Coon et al. demonstrated very efficient mean operative times of 40 min with robotic assisted TKA, which was better than conventional operative times for the studied surgeons.¹⁹ Grau et al. noted that use of a dedicated “robotic” operating room and surgical team, as well as progression through various surgical steps such as landmark registration and bony resection in a methodical and unchanging manner could contribute to significantly decreased operative times.¹⁵

The value in robotic-assisted TKA remains in the flexibility to make numerically quantified adjustments which can improve component placement and indirectly assist in accurate soft tissue balancing. Additionally, this real-time feedback on component placement, in conjunction with assessment of bone quality, plays a role when we are making the decision on whether to proceed with cemented or cementless components. In the aforementioned study by Grau et al. the average cementless RTKA took only 4.1 min less than the cemented cohort.¹⁵ Adjustments in bony resection and component positioning, as well as decisions to perform soft tissue balancing or consider utilizing bone cement, are easily made when utilizing the robot. It is our experience that multiple repetitive bone cuts and attempts to achieve soft tissue balancing, which are often seen in conventional jig assisted TKA, have been minimized with accurate robotic planning. This effectively decreases overall surgical time and allows for better component placement.

This study utilized a single robotic arm assisted device. There are several differences among contemporary robot platforms that may contribute to differences in operative times. The robotic system we used requires a saw blade and utilizes a preoperative CT scan. Registration and verification is done through checkpoints. In comparison, other robotic devices may use an intraoperative “paint” based method for registration, where the robot is used to analyze the surface of the bone intraoperatively. Other robotic systems may use a burr instead of a saw blade to make bone cuts. Further studies should be performed to elucidate the effect of various robotic devices on overall surgical efficiency.

There were several limitations to our study that must be considered when interpreting the results. First, this study analyses the results from a single fellowship-trained arthroplasty surgeon, using a single robotic system (Stryker Mako Robotic System). The learning curve and time to reach a steady state of surgical efficiency may not be directly applicable to other surgeons and surgical teams. The surgeon utilized non-cemented (“press fit” technique) total knees for 72% of cases and cement in 28%, which have been shown in other studies to decrease the time of implantation when compared to cemented total knees.¹⁵ In addition, this study included total duration of procedure, which is derived from start and procedure in the OR procedure log. However, it may be useful in future studies to characterize specific time points during surgery, such as the time from incision to prosthesis implantation. There may be a possibility that certain time points may differ from the pattern seen for total duration of procedure. Furthermore, this study solely investigates operative duration for robotic arm-assisted TKA without comparing it to a control group of conventional jig-based TKA procedures.

5 Conclusion

In this cohort of 682 consecutive robotically assisted total knee arthroplasties, the authors report substantial improvement in procedure duration between the first 50 procedures and later groups of 50, with a plateau in improvement inefficiency between the 151th and 200th procedure. No evidence of differences in length of stay, readmissions or reoperations were identified. The findings of this study will allow surgeons to better understand the implications of integrating robotic arm-assisted total knee arthroplasty into their practice. Surgeons can expect significant improvement of their operative time following completion of at least 50 robotic assisted procedures and reach a maximum level of efficiency between 151 and 200 procedures.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Ethical approval

Protocol reviewed and approved by:

St. Elizabeth Healthcare.

Institutional Review Board.

Author contributions

All authors contributed to the work of this manuscript. Each author reviewed and commented on the final paper.