Learning curve for imageless robotic-assisted total knee arthroplasty in non-fellowship trained joint replacement surgeons

Abstract

Introduction

Robotic-assisted total knee arthroplasty (RA-TKA) has become increasingly popular, although an associated learning curve can be a deterrent for some surgeons. Prior studies have addressed this learning curve in fellowship-trained arthroplasty surgeons, however the learning curve among non-fellowship-trained surgeons remains unclear. The objective of this study was to investigate the learning curve for imageless RA-TKA related to operative time and rates of complications among two non-arthroplasty-trained orthopedic surgeons.

Methods

This retrospective case series included 200 RA-TKA consecutive cases performed by two non-arthroplasty-trained orthopedic surgeons (100 each). Cases were divided into 2 cohorts for each surgeon: the first 50 consecutive cases and the second 50 cases. These cohorts were then compared to assess for trends in each surgeon as well as in both surgeons combined. Mean operative times were compared, as were hospital length of stay, complications, readmission, and reoperations.

Results

For both surgeons, the mean operative time significantly decreased from the first 50 cases to the next 50 cases (116.5 vs 108.4 min for surgeon 1, P = 0.031; 125.7 vs 109.1 min for surgeon 2, P = 0.001). No significant differences were found among length of stay, complications, readmissions, or reoperations between cohorts.

Conclusion

General orthopedic surgeons can expect to optimize operative time within 50 cases, while not carrying associated risks of related complications during the early learning period.

Abbreviations

TKA

total knee arthroplasty

RA

robotic-assisted

LOS

length of stay

SSI

surgical site infection

PJI

prosthetic joint infection

I&D

irrigation and debridement

MUA

manipulation under anesthesia

mTKA

manual total knee arthroplasty

BMI

body-mass index

1. Introduction

The incidence of total knee arthroplasty (TKA) in the United States continues to rise, with a projected 1.5 million TKAs performed annually by 2050.¹ Robotic-assisted TKA (RA-TKA) has gained popularity since its advent due to expectations that it will enhance surgeons’ accuracy, lower complication rates, and offer improved patient satisfaction scores.² Comparative studies of robotic-assisted and conventional TKA have demonstrated that RA-TKA provides improved accuracy of bone cuts and gap balancing, and decreased variability in implant positioning and postoperative mechanical alignment.3, 4, 5

Alternatively, some studies have reported no improvements over manual TKA in long-term clinical and functional outcomes up to 10 years.⁶ The use of RA-TKA has also been criticized for higher costs, longer surgical times due to an associated learning curve, and disruptions to operating room efficiency.² The learning curve required for a surgeon to become proficient with RA-TKA can result in a sharp increase in surgical time, although literature on the overall impact of RA-TKA on operative time is variable. For instance, a learning curve associated with RA-TKA has been shown to decrease, maintain, or even increase operative times compared to manual TKA (mTKA).^7,8

Although numerous studies describing the learning curve associated with RA-TKA have been published, all but one have included fellowship-trained arthroplasty surgeons. A 2022 retrospective review by Ali et al.⁸ was perhaps the sole study including non-fellowship trained surgeons, and the authors used an “image-based” robotics system. In addition, few studies have focused on “imageless” RA-TKA systems.9, 10, 11, 12, 13 Therefore, the goal of this study was to assess the learning curve for imageless RA-TKA among non-fellowship-trained orthopedic surgeons regarding operative time and clinical outcomes.

2. Methods

2.1. Study design

Retrospective case series.

2.2. Setting

This retrospective case series was granted institutional review board exemption prior to data collection. All adult patients (>18 years of age) who underwent primary RA-TKA performed by two senior orthopedic surgeons between March 2018 and December 2020 were screened for inclusion. Data was collected via chart review for all eligible patients between their initial operation and their final follow-up visit. All surgeries were performed at two hospitals within a single institution. The hospitals purchased two Smith & Nephew NAVIO systems (Smith & Nephew, Andover, TX, USA), one at each center. The NAVIO Surgical System is an example of imageless RA-TKA, which utilizes intraoperative bone surface mapping instead of preoperative computed tomography (CT) to produce a 3-dimensional model of the patient's knee. Both surgeons received identical training for the NAVIO Surgical System, which included didactic education, sawbones training, and instructor-led cadaveric training.

2.2.1. Surgeon information

Neither surgeon underwent a fellowship in adult reconstruction. Surgeon 1 is a general orthopedic surgeon who had been in practice for 15 years at the beginning of this study. Surgeon 2 is fellowship-trained in sports medicine but predominantly practices general orthopedics and has been in practice for 19 years. Both were considered “medium volume” arthroplasty surgeons (50–100 TKA cases per year)¹⁴ and neither had any prior experience with robotic arthroplasty technology.

2.2.2. Surgical technique

Both surgeons used a similar surgical technique on all patients. The surgical technique involved standard surgical preparation, tourniquet application, with sterile preparation of the NAVIO instrumentation. A standard midline incision and a medial parapatellar approach was performed in all patients. The navigation system was registered intraoperatively for each case according to the manufacturer instructions. Percutaneous or intra-incisional trackers are placed in the femur and tibia, and the bony surface anatomy is mapped using a point probe. The Journey Bi-cruciate stabilizing (BCS) II implant was used in all cases (Smith & Nephew, Memphis, TN, USA). All tibial, femoral, and patellar implants were cemented, and patellar resurfacing was performed according to surgeon preference. Surgeon 1 selectively resurfaced the patella while Surgeon 2 routinely resurfaced.

2.3. Participants

The first 100 consecutive eligible RA-TKA cases for each surgeon were included and separated into two cohorts for each surgeon: the first 50 cases and the second 50 cases. Individuals undergoing primary unilateral RA-TKA for knee osteoarthritis after failed conservative management were considered eligible for inclusion. Individuals undergoing revision TKA, bilateral TKA, or unicompartmental knee arthroplasty were excluded. Patients were seen in clinic at standard postoperative intervals up to 12 months unless otherwise indicated.

2.4. Variables

The primary outcome of interest was operative time in minutes, defined by time of incision to time until completion of skin closure. Secondary outcomes included postoperative complications related to the procedure, hospital length of stay (LOS) in days, hospital readmission within 90 days, and reoperations. Postoperative complications included knee stiffness requiring manipulation under anesthesia (MUA), superficial surgical site infection (SSI), periprosthetic joint infection (PJI), aseptic loosening, reactive synovitis, and refractory pain requiring genicular knee blocks. Reoperations included revision arthroplasty, superficial incision and drainage (I&D), and deep I&D with polyethylene exchange.

2.5. Data sources/management

Medical records were reviewed for all eligible patients and pertinent variables were recorded from physician notes, operative notes, and all available follow-up notes.

2.6. Bias

The inclusion of patients on a sequential basis exposes our sample to a degree of selection bias, although this method of sampling was key to the comparison of our data in a consecutive manner.

2.7. Study size

Study recruitment was guided by current literature that suggests RA-TKA learning curves can range from 5 to 50 cases. Given that most of the available data represents learning curves in fellowship-trained arthroplasty surgeons, we planned to include the first 100 consecutive eligible RA-TKA cases for each surgeon to allow for potential differences in non-fellowship trained arthroplasty surgeons.

2.8. Quantitative variables

All continuous data are reported as mean ± standard deviation (SD). Categorical data are reported as counts and percentages.

2.9. Statistical method

Descriptive statistics were calculated from continuous and categorical data. Normality testing was performed on continuous data, and means were compared using unpaired t tests and Mann-Whitney U tests for parametric and non-parametric data, respectively. Categroical data was analyzed using the Chi-square or Fisher's exact test. Statistical significance was based on an alpha level of 0.05.

3. Results

3.1. Participants

A total of 267 primary RA-TKA cases were identified between both surgeons. Of these, 14 were excluded for being bilateral cases performed during the same anesthetic event in seven patients. Two hundred fifty-three cases were confirmed eligible, from which the first 100 consecutive cases from each surgeon were included for a total of 200 cases.

3.2. Descriptive data

Mean patient age was 64.7 ± 8.0 years with 64% being female. Mean body-mass index (BMI) was 33.6 ± 5.7 kg/m². The left knee was operated on in 94 cases (47.0%) and the right in 106 (53.0%). Patient demographic information is shown in Table 1.

Table 1.

Patient demographics.

	All (n = 200)	Surgeon 1 (n = 100)	Surgeon 2 (n = 100)	P value
Age (years)	64.7 ± 8.0	64.1 ± 7.0	65.4 ± 8.9	0.227
Female	128 (64.0)	61 (61.0)	67 (67.0)	0.462
BMI (kg/m2)	33.6 ± 5.7	34.1 ± 5.7	33.2 ± 5.6	0.290

Values expressed as n (%) or mean ± SD. BMI, body-mass index.

3.3. Outcome data

Postoperative outcomes for both surgeons are listed in Table 2. There were 30 total complications for both surgeons. The reasons included postoperative knee stiffness requiring MUA (n = 19), superficial wound complications (n = 2), refractory pain (n = 2), PJI (n = 3), reactive synovitis (n = 2), and aseptic loosening (n = 2). Of these, 16 occurred in the first 50 cases compared to 14 in the last 50 (P = 0.843). The incidences of individual complications within the two timeframes are shown in Table 3.

Table 2.

Postoperative outcomes for both surgeons.

	All (n = 200)	First 50 cases (n = 100)	Second 50 cases (n = 100)	P value
LOS (days)	2.0 ± 1.0	2.1 ± 0.9	1.9 ± 1.0	0.164
Complications	30 (15.0)	16 (16.0)	14 (14.0)	0.843
90-day readmission	15 (7.5)	6 (6.0)	9 (9.0)	0.593
Reoperation	15 (7.5)	6 (6.0)	9 (9.0)	0.593
Revision	7 (3.5)	3 (3.0)	4 (4.0)	>0.999

Values expressed as n (%) or mean ± SD. LOS, length of stay.

Table 3.

Individual complication rates for both surgeons.

	All (n = 200)	First 50 cases (n = 100)	Second 50 cases (n = 100)	P value
Stiffness requiring MUA	19 (9.5)	10 (10.0)	9 (9.0)	>0.999
Aseptic loosening	2 (1.0)	1 (1.0)	1 (1.0)	>0.999
Superficial wound	2 (1.0)	1 (1.0)	1 (1.0)	>0.999
PJI	3 (1.5)	3 (3.0)	0 (0.0)	0.246
Refractory pain	2 (1.0)	1 (1.0)	1 (1.0)	>0.999
Synovitis	2 (1.0)	0 (0.0)	2 (2.0)	0.498

Values expressed as n (%). MUA, manipulation under anesthesia; PJI, periprosthetic joint infection.

A total of 19 cases of knee stiffness requiring MUA occurred, 10 in the first 50 cases versus 9 in the second 50 (P > 0.999). The two superficial wound complications were due to dehiscence and required superficial I&D at 38 and 52 days. Both cases of persistent postoperative pain necessitated a genicular knee block at 8 and 15 months. Among the 3 cases of PJI, two were managed with I&D and polyethylene exchange at 9 and 14 days postoperatively, and one was managed with a two-stage revision (explant and antibiotic spacer placement followed by revision) performed at 19 and 23 months. The two cases of reactive synovitis required arthroscopic synovectomy at 4 and 10 months, but both cases also went on to be revised at 15 and 22 months, respectively. Both cases of aseptic loosening resulted in revision at 12 and 41 months. In total, 7 cases required revision at an average of 14.4 months, ranging from 0 to 41 months.

3.4. Main results

3.4.1. Both surgeons

The mean operative time for all cases was 114.9 ± 22.8 min. There was a significant decrease in operative time from 121.1 ± 22.5 min for the first 100 cases (50 per surgeon) to 108.7 ± 21.6 min for the next 100 cases (P < 0.001) (Table 4). The downward trend in operation time for both surgeons is illustrated in Fig. 1. Surgeon 1 resurfaced the patella in 18% of cases (6 in the first cohort and 12 in the second), while Surgeon 2 resurfaced the patella in 100% of cases.

Table 4.

Operative time.

	First 50 cases	Second 50 cases	P value
Both surgeons (n = 200)	121.1 ± 22.5	108.7 ± 21.7	<0.001
Surgeon 1 (n = 100)	116.5 ± 17.9	108.4 ± 19.3	0.031
Surgeon 2 (n = 100)	125.7 ± 25.7	109.1 ± 23.9	0.001

Values expressed in minutes as mean ± SD.

Fig. 1.

Operation time is shown over the course of 100 cases for both surgeons. X axis displays the running case number (1–100), Y axis displays time in minutes. Dotted lines depict the linear trendlines for each surgeon.

*Fig. 1 is intended to be printed in color.

3.4.2. Surgeon 1

Surgeon 1 operated on 39 males and 61 females, with a mean age of 64.1 ± 7.0 years and a mean BMI of 34.1 ± 5.7 kg/m². No significant differences were found in patient demographics between the first 50 and second 50 cases. The mean operative time decreased significantly from 116.5 ± 17.9 min for the first 50 cases to 108.4 ± 19.3 for the second 50 cases (P = 0.031). No significant differences were found among LOS, complications, readmissions, reoperations, or revisions (Table 5).

Table 5.

Postoperative outcomes for Surgeon 1.

	All (n = 100)	First 50 cases (n = 50)	Second 50 cases (n = 50)	P value
LOS (days)	1.9 ± 1.2	2.2 ± 1.1	1.7 ± 1.3	0.052
Complications	18 (18.0)	10 (20.0)	8 (16.0)	0.806
90-day readmission	5 (5.0)	2 (4.0)	3 (6.0)	>0.999
Reoperation	11 (11.0)	5 (10.0)	6 (12.0)	>0.999
Revision	5 (5.0)	2 (4.0)	3 (6.0)	>0.999

Values expressed as n (%) or mean ± SD. LOS, length of stay.

3.4.3. Surgeon 2

Surgeon 2 operated on 33 males and 67 females, with an average age of 65.4 ± 8.9 years and a mean BMI of 33.2 ± 5.6 kg/m². There were no significant differences in patient demographics between groups. The mean operative time decreased significantly from 125.7 ± 25.7 min for the first 50 cases to 109.1 ± 23.9 for the next 50 cases (P = 0.001). There were also no significant differences among LOS, complications, readmissions, reoperations, or revisions (Table 6).

Table 6.

Postoperative outcomes for Surgeon 2.

	All (n = 100)	First 50 cases (n = 50)	Second 50 cases (n = 50)	P value
LOS (days)	2.1 ± 0.6	2.1 ± 0.4	2.2 ± 0.7	0.377
Complications	12 (12.0)	6 (12.0)	6 (12.0)	>0.999
90-day readmission	10 (10.0)	4 (8.0)	6 (12.0)	0.748
Reoperation	4 (4.0)	1 (2.0)	3 (6.0)	0.617
Revision	2 (2.0)	1 (2.0)	1 (2.0)	>0.999

Values expressed as n (%) or mean ± SD. LOS, length of stay.

4. DISCUSSION

4.1. results

This retrospective case series is the first of its kind to assess the learning curve for “imageless” robotic-assisted total knee arthroplasty in two non-arthroplasty fellowship-trained surgeons. Both surgeons had significant reductions in mean surgical time between the first and second 50 cases. Surgeon 1 reduced his average time by 8.1 min while Surgeon 2 reduced his time by 16.6 min. Neither surgeon had significant differences in complications between cohorts. Furthermore, the distribution of complication type did not vary between cohorts, and it is unlikely that any complications were directly related to the use of robotic-assisted technology. Our findings suggest that proficiency of navigation-based RA-TKA can be accomplished by the 50th case without an increased risk for complications during the early learning phase.

RA-TKA has become a staple in the arthroplasty community thanks to its advantages in perioperative planning, operative precision, and reduced soft tissue damage.⁸ However, RA-TKA also raises some concerns for new adopters, such as longer operative times, higher costs, and an associated learning curve.¹⁴ Among studies that compared surgical times between manual TKA and robotic-assisted arthroplasty, RA-TKA took significantly longer than mTKA, but the difference became less significant after 10–20 cases, on average.^12,15, 16, 17, 18 Current literature suggests RA-TKA learning curves can range from 5 to 50 cases, although methods of measurement vary between studies. For example, some studies compared operative time on a case-by-case basis, noting significant reductions as early as the 7th case.^7,12 While other studies, like ours, looked at average operative times in grouped case increments, with inflection points cited after the 25th and 50th cases.^13,19

Learning curves are largely dependent on the individual, although numerous surgeon-specific characteristics have been associated with learning curve duration. For instance, Vermue et al.¹⁵ pointed out the impact of annual TKA case volume. These authors reviewed consecutive cases performed by three high-volume surgeons (>100 TKA cases per year), one medium-volume (50–100), and two low-volume (<50) surgeons through their first 17 months using a RA-TKA system, and found that only the high-volume surgeons had significant reductions in surgical time.¹⁵ Prior experience with robotic-assisted or computer-navigated systems is another contributing factor that was highlighted by Schopper et al.²⁰ This study found that when a surgeon without prior robotic experience routinely performed RA-TKAs with the assistance of another surgeon who was highly experienced with robotic-assisted arthroplasty, the surgical times were significantly shorter compared to another novice surgeon who only received assistance from the experienced surgeon on occasion. The authors concluded by suggesting that manufacturers should provide expanded assistance to centers without prior experience using robotic-assisted technology.²⁰

As the demand for TKA in everyday surgical practice continues to rise, the number of cases performed by non-arthroplasty fellowship-trained surgeons will undoubtedly rise as well.¹⁴ Non-fellowship trained surgeons tend to have increased operative times and complication rates compared to their arthroplasty fellowship-trained colleagues,²¹ thus it is important to consider the implication of new technologies for these surgeons. Ali et al.⁸ were the first to look at the learning curve for RA-TKA in general orthopedic surgeons. They reviewed the first 60 consecutive image-based RA-TKA cases in two non-fellowship-trained orthopedic surgeons practicing in a high-volume community setting. Despite their finding that one surgeon did not display a difference in surgical time between the first and last case, they concluded that a reduction in operative time could be achieved within the first 40 cases.⁸ Similarly, our findings suggest that a significant reduction in time can occur within 50 cases for the imageless RA-TKA. A recent systematic review¹⁴ of RA-TKA learning curves found the average inflection point for operative time to be 14.9 cases, which is substantially shorter than our results. However, nearly all of the surgeons included in this review were fellowship trained in arthroplasty, thus our findings highlight a potential difference in learning curve duration between fellowship and non-fellowship trained surgeons. It is worth noting, however, that both Ali et al. and this current study compared average operative times among cohorts of 20 and 50 cases, respectively, which may exaggerate the inflection point for non-arthroplasty fellowship-trained surgeons.

Multiple studies have assessed learning curves specific to the NAVIO imageless RA-TKA system with varying results.9, 10, 11, 12, 13 In a review of the first 72 imageless RA-TKA cases in a single surgeon, Collins et al.¹⁰ failed to identify a substantial learning curve other than a nonsignificant 2% reduction of operative duration per month. However, the surgeon in this study had prior experience with other computer-navigated arthroplasty systems, which has been shown to flatten the learning curve.²⁰ A similar review by Vaidya et al.¹³ assessed the first 75 imageless RA-TKA cases in an arthroplasty-trained surgeon and concluded an approximate learning curve of 25 cases. A longer learning curve for the imageless platform could be explained by the lack of detailed preoperative planning that image-based systems provide. Instead, surgeons must register the patient by hand, which relies on the surgeon's ability to accurately map the patient's anatomy using point probes at the time of surgery.² By eliminating the need for preoperative advanced imaging, however, imageless systems provide patients the benefits of reduced costs, increased convenience, and less radiation exposure.² The other potential increase in time may be related to the use of a robotic burr which is inherently slower than a robotic saw as seen with other robotic systems.

The implications of our findings are significant for surgeons without experience using robotic arthroplasty systems who are interested in incorporating imageless RA-TKA into their practice. We combined results from two medium-volume TKA surgeons, both without prior experience using robotic technology, which we believe expands the generalizability of our findings. This study should reassure all orthopedic surgeons, and especially non-fellowship-trained general orthopedic surgeons, that an imageless RA-TKA system can be safely adopted within a reasonable timeframe.

4.2. Limitations & strength

This study contains several limitations. The operative times and rates of complications were not compared to the surgeons’ outcomes with mTKA, so it remains uncertain how these outcomes compare to their baselines. While sparse, the exclusion of bilateral TKAs and UKAs performed on the robotic system may have provided extra experience to these surgeons that was not directly reflected in the learning curve demonstrated in this study. Unlike other studies, we did not assess the learning curve for accuracy of intraoperative and postoperative measurements (implant positioning, gap balancing, limb alignment). Furthermore, our results are limited to the experiences of two surgeons, and while we believe their differences contribute to the overall generalizability of our findings, we acknowledge that two surgeons are not representative of the larger population. Despite these limitations, our results address a lack of literature on the learning curve for “imageless” RA-TKA among non-arthroplasty fellowship-trained surgeons. Future studies should aim to identify which steps during the imageless RA-TKA procedure contribute most to operative times for general orthopedic surgeons, so that adjustments can be made in surgical workflow and by manufacturers to further reduce the learning curve.

5. Conclusion

Robotic-assisted total knee arthroplasty is here to stay, and while its implementation may pose a challenge, our study suggests that reduced operative times can be achieved within the first 50 cases. Our findings provide evidence that non-arthroplasty fellowship-trained surgeons may have a longer learning curve than their arthroplasty fellowship-trained counterparts, however, these surgeons can still reach an optimized operative pace without placing patients at higher risk for complications during the early learning period. Therefore, general orthopedists should not be deterred from adopting robotic-assisted imageless technology in their practice. Instead, they should consider it as an asset that can potentially improve surgical outcomes and patient satisfaction.

Funding

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

Ethical approval and patient consent

This study received institutional review board exemption for human subject research with less than minimal risk. A waiver of consent and HIPAA authorization was obtained for the retrospective collection of patient information.

Declaration of patient consent form

Not applicable.

Financial support and sponsorship

None.

Funding

None.

CRediT authorship contribution statement

Samuel D. Stegelmann: Conceptualization, Data curation, Formal analysis, Writing – original draft, Writing – review & editing, Visualization. Justin Butler: Conceptualization, Writing – review & editing, Supervision. Samuel G. Eaddy: Writing – original draft, Writing – review & editing, Visualization. Trent Davis: Data curation, Writing – review & editing. Kirk Davis: Conceptualization, Supervision, Project administration. Richard Miller: Conceptualization, Supervision, Project administration.

Declaration of competing interest

None.