Introduction

Total Knee Arthroplasty (TKA) is considered the gold standard treatment for end-stage knee osteoarthritis [1, 2]. Limb alignment and implant positioning are considered among the most important prognostic factors for long-term implant survivorship, patient satisfaction and clinical outcomes [1, 3, 4]. To improve that, computer navigation and robotic-assisted surgery have gained interest as tools that help the surgeon to perform TKA with greater precision and accuracy.

Robotic technologies have been introduced with the expectation to improve component positioning providing increased accuracy and consistency in bone resections while reducing soft tissue damages [5, 6]. Instability, aseptic loosening, malalignment, and increased polyethylene wear are among the most frequent causes of implant failure, and all of them are related to a poorly balanced and malaligned knee [7]. In addition, components malpositioning is frequently associated with patient dissatisfaction, patellofemoral complications, and it may leads to implant failure and early revision surgery [8, 9]. Robotic surgeries have proven successful in increasing early functional outcomes and reducing radiographic outliers compared with conventional technique, however, mid-to-long term studies have yet to prove meaningful enhancement in terms of implant survivorship, patient satisfaction, and functional outcomes [10]. However, most of the available literature refers to CT-guided robotic TKA [11,12,13], while few studies report on imageless robotic systems [14, 15]. In fact, avoiding the need of a preoperative CT-scan, imageless robotic systems have the advantages of reducing the radiation exposition of the patients while ensuring smoother preoperative and intraoperative procedures.

Imageless navigation is a simple solution, relatively recently introduced as a tool to drive the distal femoral and proximal tibial cuts, to facilitate and improve the precision of TKAs’ coronal alignment, avoiding intramedullary canal guides, when aiming for a mechanical axis of the lower limb [16].

Different studies have been published on potential functional and alignment advantages of robotic-assisted TKA (RTKA) compared to conventional manual TKA, but to the best of the authors’ knowledge, there is a lack of evidence on comparison between RTKA and navigated TKA (NTKA). The primary endpoint of this study was to compare the 1-year clinical outcomes following imageless RTKA and imageless NTKA. Our hypothesis was that RTKA was associated with improved patient-reported outcome measures (PROMs) and range of motion (ROM) at 1-year follow-up.

Materials and methods

Patients that underwent total knee arthroplasty (TKA) for primary knee osteoarthritis between February and September 2020 were prospectively followed and retrospectively evaluated. All surgical procedures were performed by experienced knee arthroplasty surgeons at a high-volume institution. After excluding the first 25 robotic cases to allow for an appropriate learning curve window for the surgical equipe, patients’ data were collected and analyzed. During this time, a total of 173 primary TKAs were performed in the orthopaedic department, including 87 consecutive navigated TKAs (imageless NTKA, iAssist Knee, Zimmer, Warsaw, IN) and 61 consecutive robotic TKAs (RTKA, ROSA Knee System, Zimmer Biomet, Warsaw, IN) using a single implant design (Persona PS Knee, Zimmer, Warsaw, IN). Eleven patients in the RTKA group were excluded from this cohort for different reasons such as: patients with ligament insufficiency that required higher level of constraint, patients with a diagnosis other than primary osteoarthritis, patients with a deformity requiring augmentation, neurological movement disorders, patients with severe varus deformity (> 15°), patients with valgus deformity, and patients that did not complete their 1-year clinical and radiographic follow-up. Fifty RTKA patients with mild-to-moderate varus deformity (≤ 15°) undergoing primary unilateral TKA were, finally, included for final analysis with a mean follow-up of 13.4 months (± 1.3, range 12–15). Varus deformity was defined as a hip–knee–ankle (HKA) angle greater than or equal to 1° with the mechanical axis (MA) passing medial to the knee joint. A matched cohort of 47 consecutive navigated TKAs performed during the same period, was used for comparison with a mean follow-up of 13.6 months (± 1.2, range 12–15).

The electronic medical records were collected prospectively and reviewed retrospectively for preoperative basic demographic information, intraoperative events and postoperative function and complications. Demographic information included age (years), weight (kg), height (m), sex, body mass index (BMI, kg/m2), and operated side. Intraoperative information included operating room (OR) time (min). OR time was considered from when the patient entered the theatre to when he/she left it, including robotic device setup time. Surgical anesthesia was performed before entering the theatre. Complications were classified as wound complications, including cellulitis or wound drainage of the surgical site or the pin insertion site (treated conservatively), and reoperation for any cause (debridement, antibiotics and implant retention [DAIR], two-stage revision for periprosthetic joint infection [PJI], and revision for any other cause). If a patient with a superficial wound infection was later found to have a PJI, he/she was excluded from the wound complication count and only included in the reoperation group.

Patients were followed-up clinically and radiographically at 3, 6, and 12 months postoperatively. Full medical records were reviewed, including preoperative Knee Society Knee (KSSk) and Function score (KSSf), range of motion (ROM), and lower limb alignment (hip–knee–ankle angle [HKA]). Knee function at postoperative evaluation was assessed at 12-month follow-up with the KSSk score, KSSf score, Knee injury and Osteoarthritis Outcome Score (KOOS) score, and Forgotten Joint Score (FJS-12) [17, 18]. Clinical scores were collected by a trained research-fellow at 1-year follow-up. Radiographic analysis of preoperative and postoperative images evaluating HKA angle was performed by independent fellowship-trained orthopedic surgeon [19, 20]. Radiographic outliers were considered when falling outside of a cutoff of ± 3° from MA.

All surgical procedures were performed through a standard medial parapatellar approach, and the same prosthetic implant was used in every case (Persona Posterior Stabilized [PS] Knee, Zimmer, Warsaw, IN). The ROSA Knee System (Zimmer Biomet, Warsaw, IN) is a collaborative robotic device that allows the surgeon to proceed with an image-based planning using a 3D virtual model based on specific preoperative 2D radiographs, or an image-less planning based on intraoperative bony landmarks collection and ligament balancing evaluation. All cases included in this study were performed with the image-less procedure. The surgical technique of ROSA Knee System has been previously described [21] and its accuracy demonstrated both in vitro and in vivo [22]. Preoperatively, the robotic-assisted procedure aimed for a functional personalized aliment with an under correction of the original deformity leaving 2° of varus. The aimed femoral coronal angle was 90° (LDFA), the femoral sagittal angle (FSA) was 3° according to the implant’s surgical technique, the tibial coronal angle was 88° (MPTA), and the tibial sagittal angle was 3° (TSA) according to the implant’s surgical technique.

The iASSIST (Zimmer, Warsaw, IN, US) navigation device is a handheld computer‐assisted accelerometer-based stereotaxic system integrated into a small pod which attaches to the femoral and tibial resection cutting jigs, guiding the resection at the appropriate angles in both the coronal and sagittal planes. After bone resection, the accuracy of the alignment can be validated, confirming the position of both the femoral and tibial components. The surgical workflow follows the classic method of femoral and tibial bone resection with each bone resected independently along the mechanical axis [16]. Preoperatively, the navigated assisted procedures aimed to achieve a neutral MA.

All patients in both groups followed the same standardized postoperative rehabilitation protocol. However, even though specific differences were addressed to each patient, typically rehabilitation started the day of surgery with active and passive ROM and partial weight bearing as tolerated. Stretching exercises and a light strength training program were started to reinforce the surrounding muscles. Each physiotherapy session was performed by the same multidisciplinary team in both groups. Discharge was between 4 and 7 days (mean, 5 days) after pain control, knee flexion to minimum 90°, independent mobilization with the use of crutches, and independent ascent and descent of stairs was achieved. All patients completed the above protocol with a combination of outpatient and at-home physical therapy.

The study is performed in accordance with the ethical standards in the 1964 Declaration of Helsinki and with the HIPAA regulation. The Institutional Review Board (IRB) of the author’s institution defined this study as exempt from IRB approval (prospective study on a well-established surgical procedure and commercialized implant).

Statistical analysis

A power analysis was performed using a two-sided test at an alpha level of 0.05 with a power or 80% to determine the required sample size for statistical significance according to our main endpoints with a medium effect size. Based on the power analysis, 90 patients (45 in each group) were needed to detect a significant difference. Considering a potential 5–10% dropout, the outcomes of 98 patients were analyzed. Baseline patients’ characteristics and functional scores were compared between the two cohorts with univariate analyses. Data were described using frequencies for categorical data and mean (SD) and range as appropriate for continuous data. Wilcoxon Rank Sum Test was performed for continuous variables to compare means, while Fisher Exact Test was performed for categorical variables. Histograms were created to compare the distribution of KSSk, KSSf, FJS, and KOOS score between RTKA and NTKA cohorts. ROM, HKA angle, KSS, KSSf, FJS, and KOOS were compared between the two groups using a Wilcoxon Rank Sum Test. Multivariate general linear models with baseline characteristics as covariates were used to evaluate the associations of the surgical technique with functional outcomes using the F Test based on Type III sum squares in the ANCOVA model. The covariates analyzed were surgical technique (NTKA vs RTKA), age, sex, and BMI. All these variables were included in the regression model, and variables with the weakest correlations were sequentially removed in subsequent models until that with the strongest association was identified.

All data were analyzed using SAS 9.4 (SAS Institute, North Carolina, US) and statistical significance was maintained at a p value of less than 0.05.

Results

Demographics and perioperative outcomes

There were no significant differences in age, gender, and BMI between the two groups (Table 1). Compared to the NTKA group, the RTKA group reported a statistically significant longer surgical duration (122 min vs. 97 min; p < 0.0001), while no significant difference was reported for mean recovery days (5.1 days vs. 4.9 days; p = 0.2975) (Table 2, Fig. 1).

Table 1 Baseline characteristics of robotic and navigated patients
Full size table
Table 2 Perioperative and intraoperative evaluation
Full size table
Fig. 1
figure 1

OR time comparison between the two groups

Full size image

PROMs

There were no significant differences in preoperative KSSk and KSSf score and postoperative FJS and overall KOOS score. Significant differences were reported for the “Pain” (p = 0.0283) and “Function” (p = 0.0360) subsection of the KOOS score, however, the model F test p values showed that “age” was an influencing cofounding factor for the “Function” subsection (F Test p = 0.1087), and after having covariate age in the model, treatment impact to KOOS Function score was not significant anymore (F Test p = 0.0550) (Fig. 2). Postoperative KSS knee and function scores were significantly higher in the RTKA group (p < 0.0001), however, F test p values based on Type III sum squares in the ANCOVA model indicated that treatment (p < 0.0001) and gender (p = 0.0226) had significant impact on subjects’ KSS Knee Score. The interaction between gender and treatment arm were not significant at level of 0.05 (p = 0.3000). In addition, F test based on Type III sum squares indicated that both BMI (p = 0.0381) and treatment (p < 0.0001) had significant impact on KSS Function Score (Table 3).

Fig. 2
figure 2

KOOS pain and function distribution

Full size image
Table 3 Patient reported outcomes (PROMs)
Full size table

Range of motion

Preoperative ROM was not significantly different between the two groups. F test p value based on Type III sum squares in the ANCOVA model showed that treatment had significant impact on ROM at final follow-up (p < 0.0001). Comparing least square means from general linear model, RTKA group had higher maximum range of motion compared to NTKA group (119.4° vs. 107.1°; p < 0.0001), and the mean delta improvement in the arc of motion from preoperative was significantly higher in the RTKA group at the last follow-up by 11.67° (23.02° vs. 11.36°; p < 0.0001) (Table 4).

Table 4 Range of motion comparison between RTKA and NTKA
Full size table

Radiographic outcomes

No significant difference was reported on the preoperative and postoperative HKA angle between the two groups; however, the overall number of postoperative outliers was higher in the RTKA group (30.00% vs. 17.02%; p = 0.1570). (Table 5, Fig. 3).

Table 5 Overall lower limb alignment (HKA angle)
Full size table
Fig. 3
figure 3

Postoperative HKA distribution

Full size image

The mean postoperative HKA angle in the RTKA group was 177.42° (SD ± 1.7°), the mean femoral coronal alignment based on LDFA was 89.77° (SD ± 1.4°), the mean FSA was 87.47° (SD ± 1.5°), the mean MPTA was 87.99° (SD ± 0.9°), and the mean TSA was 86.29° (SD ± 1.0°). Considering the preoperatively aimed angles, the mean HKA deviation from target was 1.5° (range, 4.2°–0.1°), the mean LDFA deviation from target was 1.3° (range, 3.6°–0.1°), the mean FSA deviation from target was 1.3° (range, 3.8°–0.1°), the mean MPTA deviation from target was 0.7° (range, 2.3°–0.1°), the mean TSA deviation from target was 0.9° (range, 3.2°–0.1°).

Complications and reoperations

Overall, one case required reoperation in the RTKA group (2%) compared with three cases in the NTKA group (6.4%, n.s.) (Table 6). Wound complications were reported in two cases in the RTKA group (4%) as compared to four cases in the NTKA group (8.7%, n.s.). None of the wound complications were directly related to the specific devices used and none of them were at pin sites used for intraoperative planning and reference. No cases of periprosthetic fractures or fractures at pin sites were reported.

Table 6 Medical events and complications
Full size table

Discussion

The most important findings of this study were that at 12-month follow-up the robotic-arm assisted imageless TKA reported better improvement in ROM and KOOS pain score compared to navigated TKA suggesting favorable functional outcomes at short-term follow-up. This is in line with what recently reported in the literature [10], however, there is a lack of evidence on clinical outcomes of imageless RTKA compared to the conventional technique, while numerous studies have described the outcomes of CT-based robotic systems. To the best of authors’ knowledge, this is the first study that compares the ROSA Knee System with an imageless navigation system instead of a conventional technique.

This study reports a significant mean OR time difference of 25 min between the two groups (122 min RTKA vs 97 min NTKA). As previously stated, we intended the OR time from the moment the patient entered the surgical theatre and not only from skin incision to wound closure. Clearly, time for surgical anesthesia was not included since it was performed outside the theatre. For this reason, our results contrast with the current literature on robotic surgery ranging from 70 to 90 min [23,24,25,26], where OR time is usually intended from skin incision to wound closure. The difference between the two groups is probably related to the time necessary to prepare and drape the robot, placing and removing the tracking pins, positioning of the camera, acquisition of the bony landmarks, and the intraoperative planning. However, as it is shown in the graph (Fig. 1), in some cases, the surgical procedure lasted 95 min as it was for the majority of the navigated TKAs suggesting that with a trained surgical team, it is possible to reduce remarkably the time of the additional steps required for the robotic technique. In addition, it is worth pointing out that in this study the surgical team was not always the same, enhancing the fact that an untrained team may slow down the procedure. Although, the surgical time registered were in line with what reported in the literature especially with the addition of newer techniques aiming to increase the accuracy and precision of the surgery as reported by Shah et al. [27].

A significant difference in total range of motion (119.4° vs. 107.1°; p < 0.0001) was detected as well as a mean delta improvement from preoperative baseline condition between RTKA and NTKA (23.02° vs. 11.36°; p < 0.0001). In addition, despite the detected differences on KSS and the overall KOOS being not significant, a significant improvement in the “pain” subsection of the KOOS in the RTKA group was reported compared to the NTKA (p = 0.0283). Similarly, Marchand et al. [28] reported the clinical outcomes of 53 RTKA compared to 53 manual TKA and noted that the RTKA cohort had significantly improved mean Western Ontario and McMaster Universities (WOMAC) total (p = 0.03) and physical function scores (p = 0.02) when compared with the manual cohort. Furthermore, the authors reported that the mean pain score was also lower in the RTKA group, however, without reaching statistical significance (p = 0.06). Moreover, in the recent meta-analysis by Agarwal et al. [10], the authors reported a significant clinical improvement using the Hospital for Special Surgery (HSS) and WOMAC scores when the robotic-assisted technique was used compared to the conventional one. However, none of the studies included referred to robotic devices as the ROSA Knee System, but all of them were CT-based instead. Nevertheless, despite several studies included in the review reported significant improvement in pain perception and ROM, it was unclear to the authors whether the differences in PROMs resulted in a clinically significant differences in pain reduction and improved function between robotic and conventional arthroplasty techniques. Conversely, Held et al. [29] compared the outcomes of 111 imageless RTKA with 110 conventional TKA at 12- and 24-month follow-up and reported no significant differences between the two groups in terms of PROMs and total ROM. However, the authors used a different handheld imageless robotic device (NAVIO™, Smith & Nephew, Memphis, US). The difference detected in this study may be related to a personalized component positioning and joint line restoration of the robotically assisted technique, however, larger prospective randomized investigations are necessary to eventually confirm these preliminary results on imageless RTKA.

This study reported no significant differences in postoperative alignment of the knee between the two groups (HKA angle), however, the overall number of outliers was higher in the RTKA group compared to the NTKA (30% vs 17%). The proportion of outliers of the NTKA group is in line with what currently reported in the literature ranging from 5 to 20% [14, 23, 26, 30,31,32,33]. However, the number of outliers in the RTKA group contrast with what has been lately reported on robotic TKA, even though most of the studies that analyzed postoperative alignment used a different robotic device and were aiming for a MA [10, 14, 23]. This could be explained considering that the surgical technique of iAssist aims to a pure MA technique, while the ROSA-assisted knee is based on the concept of personalized TKA recreating the patient anatomy and restoring the joint line orientation allowing a more precise placement of the femoral and tibial component in the coronal plane. Moreover, the targeted alignment with ROSA in mild-to-moderate varus knees was an undercorrection to 2° varus. The MPTA and LDFA were accurate with the targeted angles, with a mean deviation of 0.8° and 1.3°, respectively. The standard deviations revealed very small distribution widths of ± 0.6° and ± 0.9°, respectively. The HKA was also precise to within 1.5° ± 1.1°, and the TSA within an error of 0.9° ± 0.7°. These results are consistent with the findings reported in the current literature for robotic-assisted TKA [23, 24, 26, 34,35,36,37], despite the limited data available on imageless robotic systems. However, further studies are necessary to clarify if this personalized alignment is responsible for the improved clinical outcomes and ROM and if there is a significative difference for the component placements in all three planes.

Regarding complications, one reoperation for acute infection was reported (2%) and two cases of wound complications not related to pin sites (4%) in the RTKA group, in line with what reported in the current literature regarding similar robotic devices with a mean complication rate between 0 and 5%, including NAVIO® Surgical System (Smith and Nephew, Memphis, USA) [23, 26, 38] and MAKO robotic system (Stryker, Mahwah, NJ) [25, 39]. No cases of pin sites infections and pin related fractures were reported, and no significant difference was reported between the two groups regarding surgical site infection. According to the current literature, periprosthetic knee fracture pins related have been reported with a rate ranging from 0.06 to 4.8% in case of robotic-assisted UKA [29], and 0.16% in computer-assisted TKA [40]. The authors reported that a transcortical pin trajectory, large pin diameter (> 4 mm), diaphyseal fixation, multiple placement attempts, and the use of non-self-drilling, non-self-tapping pins were the most commonly reported risk factors. In addition, it has been reported an incidence of pin site superficial infection between 0.47 and 0.6% [41, 42]. Despite the promising results, it needs to be mentioned that the two cohorts analyzed were formed by a small number of patients suggesting that such rare complication may not be clearly detected in so small groups. Considering that each robotic device has specific pins and trackers with clear differences among them in terms of size and distance between the pins, we believe that further studies focused on only one robotic device with larger cohorts are necessary to clearly establish the incidence of pin related complications. However, according to the results of this study, the use of ROSA knee system was not associated with increased incidence of complications compared to other robotic devices currently available suggesting its safeness and reliability.

This study has several limitations. First, as a retrospective study, although patients were prospectively followed, its nature makes it susceptible to selection and detection bias since patient were not randomized and postoperative evaluation was not blinded. Second, the study has a relatively short follow-up, and these patients, therefore, still need to be followed to evaluate survivorship, revision rates, and patient satisfaction. However, a recent study reported no different outcomes at 12- and 24-month follow-up after TKA suggesting that although the longer follow-up periods are clearly necessary, there may not be of additional value in collecting short-term outcomes in routine practice at both 1 and 2 years [43]. Third, the clinical scores used to evaluate the postoperative outcomes are not free from controversies. In fact, despite the KOOS scale has been defined more effective than the KSS-F scale in measuring patient outcomes following TKA allowing the use of a smaller cohort to show a significant difference and being more sensitive in measuring improvements after an intervention, the “pain” subsection is characterized by a higher “ceiling effect” that decreases validity [44]. Fourth, considering the nature of the personalized approach allowed by the ROSA knee system where the final alignment is chosen by the surgeon according to its personal preferences and experience, the fact that the surgical procedures have been performed by different surgeons, despite their experience, may lead to a potential performance bias. However, considering the recent diffusion of this innovative device and the lack of studies currently available, we had to include the patients of the two surgeons to have a sufficient group with minimum 12-month follow-up.

Conclusion

In conclusion, at minimum 12-month follow-up, the imageless ROSA knee system proved comparable safety with NTKA including similar length of stay, overall alignment, complication, reoperation, and revision rates despite the longer operative time. In addition, the robotically assisted technique showed higher overall ROM and improvement from preoperative condition with increased PROMs compared to the NTKA group. However, high-quality randomized control trials are necessary to support these preliminary findings and support the broad adoption of this technology.