How accurate are arthroplasty surgeons in visually estimating extension and flexion gaps in total knee arthroplasty?

Abstract

Aims

The ability of a surgeon to provide accurate visual estimates of intraoperative gaps during total knee arthroplasty (TKA) is not well understood. This study evaluated: 1) the accuracy of gap estimation in extension and in flexion; 2) the accuracy of gap estimation in the medial and lateral compartments, also in extension and flexion; 3) the differences in accuracy among surgeons; and 4) the frequency of clinically significant errors in gap estimation, defined as greater than 1 mm.

Methods

A posterior stabilized TKA was performed on seven cadaveric knees. Five fellowship-trained arthroplasty surgeons and one orthopaedic resident manually stressed each knee, and visually assessed the medial and lateral gaps in full extension and 90° of flexion. Gaps were objectively measured via a motion capture system. Gap estimation error was calculated as the difference between the surgeons’ visual assessment and the measured gaps.

Results

Across all surgeons and knees, the mean gap estimation error was -0.4 mm (SD 0.7), with the majority (72%) of gaps being underestimated. Errors were greater in extension (-0.7 mm (SD 0.8)) than in flexion (-0.2 mm (SD 1.0)) (p < 0.001). Lateral gap error was less in flexion (-0.1 mm (SD 1.0)) than extension (-0.7 mm (SD 0.8)). Gap estimation error pooled for all assessments differed between surgeons, ranging from a mean error of -0.8 mm (SD 0.8) to 0.2 mm (SD 1.2) (p < 0.001). Clinically significant gap estimation errors (> 1 mm) occurred in 33% of assessments in extension and 26% in flexion (p = 0.315, not statistically different). The frequency of such errors varied by surgeon ranging from 18% to 42% (p = 0.370).

Conclusion

Surgeons tend to underestimate intraoperative gaps during TKA, particularly in extension. Clinically meaningful gap estimation errors (> 1 mm) occurred in up to 33% (26/78) of exams, supporting the need to enhance gap assessment accuracy.

Cite this article: Bone Jt Open 2026;7(3):417–424.

Introduction

While overall patient outcomes following total knee arthroplasty (TKA) are favourable, up to 20% of patients report dissatisfaction after the procedure.^1-3 Knee instability is a leading cause of this dissatisfaction, and a common indication for revision surgery.^1-4 Accordingly, achieving intraoperative knee stability through careful gap assessment by the surgeon plays a key role in postoperative satisfaction.^5-10 Notably, achieving specific intraoperative gap ranges has been associated with improved function and reduced pain postoperatively.^9,10

Surgeons typically conduct intraoperative gap assessment to achieve specific gap ranges by manually applying medial and lateral forces to the foot and ankle, which causes varus and valgus moments at the knee.¹¹ While the knee is under load, surgeons visualize the medial and lateral gaps at the joint line, subjectively judge knee balance, and then select the thickness of the polyethylene liner.¹¹ Importantly, a close relationship exists between intraoperative gap assessment, polyethylene insert selection and the resulting ligament tensioning, and knee stability.^11,12 Therefore, a surgeon’s ability to accurately assess the medial and lateral gaps is a key part of the operation, and can distinguish patients who experience suboptimal clinical outcomes from those who do not.^9-11

Despite its importance to postoperative patient satisfaction,^9,10 intraoperative gap assessment during manual TKA remains largely feel-based and qualitative, and relies on surgeon preference and experience.¹³ Moreover, conventional manual instrumentation is still used in the vast majority of TKAs in the USA and worldwide.¹⁴ Manual TKA, however, has at least two potential limitations. The first is heterogeneity in the magnitude of force applied by the surgeon during the exam. For example, previous work revealed that the magnitude of force applied to the knee varied between surgeons when assessing gaps in mid-flexion and flexion, and the applied force was associated with preferred polyethylene insert thickness.¹¹ The second limitation and potential source of error is inaccuracy in subjectively estimating the gaps in millimetres between the femoral and tibial component. Notably, gap ranges associated with improved outcomes can be as narrow as 3 mm.^9,10 Thus, inaccuracy in visual gap assessment could contribute to heterogeneity in the resulting knee stability and suboptimal clinical outcomes. However, the accuracy of surgeons in quantifying the medial and lateral gaps in both extension and flexion during TKA remains unclear.

Therefore, the purpose of this study was to quantify surgeon accuracy in estimating gaps during TKA, and to compare the visual estimated gaps to those quantified via motion capture technology. We posed the following research questions. First, what is the accuracy of gap estimation of a cohort of surgeons? Second, does gap estimation error differ based on knee flexion angle (extension versus flexion) and compartment (medial versus lateral)? Third, does gap estimation error differ from surgeon to surgeon, across all gap estimates pooled together, and by specific compartment? Fourth, what proportion of gap estimation errors exceeds a clinically significant magnitude (> 1 mm), first in each compartment in flexion and extension, and second, by surgeon?

Methods

Study sample

After obtaining approval from the Institutional Review Board at the Hospital for Special Surgery (IRB 2016-0071-CR1), seven pelvis-to-toe cadavers were used.¹¹ Exclusion criteria included arthritis (assessed via radiographs before procurement), knee deformity, contracture, or evidence of prior injury. The demographic and implant sizing details of the seven cadavers are listed in Table I. The seven cadaveric lower limbs underwent a CT scan (Inveon; Siemens, Germany) prior to TKA to develop a 3D rendering of each femur and tibia, as described previously.¹¹

Table I.

Demographic and total knee arthroplasty implant data for each cadaver used in the study.

Characteristic	Cadaver number
	1	2	3	4	5	6	7
Age, yrs	76	68	68	83	73	64	79
Sex	Male	Female	Female	Female	Male	Female	Female
Race	White	White	White	White	White	Black	White
Weight, lbs	148	119	180	139	158	163	140
BMI, kg/m2	21.2	23.2	31.9	24.6	25.5	29.8	24.8
Femoral component size	9	6	8	5	11	6	6
Tibial component size	E	C	E	C	F	C	D

TKA technique and 3D modelling of the femur and tibia

For all cadavers, a single experienced arthroplasty surgeon (PKS) performed a standard midline TKA exposure via a medial parapatellar approach on the right knee. The anterior and posterior cruciate ligaments were released. Bone cuts were performed with manual instrumentation using measured resection technique and mechanical alignment principles. Each specimen was implanted with a posterior stabilized, fixed-bearing TKA trial implant (Zimmer Persona System, Zimmer Biomet; USA). At the conclusion of the TKA procedure, four clusters of reflective markers (each consisting of three 9 mm diameter spheres) were rigidly fixed into the anterior aspects of the femur (two clusters) and of the tibia (two clusters) at 10 cm from the joint line for motion capture. Next, all cadaveric legs underwent a second post-procedural CT scan of the right femur and tibia. Volumetric renderings of the right tibia and femur were obtained via segmentation and image reconstruction (Mimics Research 22.0; Materialise, USA).

Qualitative and quantitative gap estimations

Independent assessments of ligament balance on the operated cadaveric knees were conducted by six individuals: five fellowship-trained arthroplasty surgeons (PSK, DJM, MBC, ADP, and GHW) with varying levels of experience (range of 6 to 24 years in practice) and a chief resident (CAK) (Figure 1). At the discretion of each surgeon, a polyethylene insert of desired thickness and constraint was selected. Regarding constraint, the surgeons had the option of inserting a standard posterior-stabilized or a mid-level constrained liner. Insert thickness was available in 1 mm increments from 10 to 14 mm, and 2 mm increments from 16 to 20 mm. After selecting the insert, each surgeon manually applied varus and valgus loads at extension and 90° (flexion), respectively (Supplementary Material). At each of these positions, they visually estimated the medial and lateral joint gaps (mm). The selected insert thickness, as well as the surgeon’s medial and lateral gap estimates in both extension and flexion, were recorded for each assessment.

Fig. 1.

A surgeon examining knee balance on a human cadaveric leg by applying varus and valgus moments at A) extension (~0° of flexion) and B) flexion (~90° of flexion). The surgeon visualized the lateral and medial gaps at the knee joint.

During each surgeon’s evaluation, knee laxity data was captured using a 3D motion capture system (MoCap System; Motion Analysis Corporation; USA). The motion capture data and the 3D reconstructed femur and tibia were integrated using dynamic analysis software (ADAMS v19, MSC Software Corp, USA) to calculate the medial and lateral gaps during the surgeons’ evaluations as described previously.¹¹

The accuracy of the motion capture system was quantified using rigid foam models of a femur and a tibia (Sawbones, Inc, USA), which were spanned with rubber bands to simulate the restraint provided by the collateral ligaments to represent a rudimentary knee model. Rigid marker clusters were affixed to the foam models of each bone. Subsequently, the tibia was manually displaced in both valgus and varus relative to the femur to simulate medial and lateral gaps observed intraoperatively. This manual positioning was conducted at several flexion angles including extension and flexion. Knee gaps were measured with a digital caliper (resolution ± 0.02 mm, General UltraTech, China) and compared to the corresponding gaps calculated via motion capture yielding a measurement accuracy of < 0.1 mm.¹¹

Gap estimation error

The primary outcome of the study was the gap estimation error, defined as the difference between the estimated gaps (via surgeon visual inspection) and the measured gaps (via motion capture). Gaps measured using motion capture were considered the gold standard. Negative and positive errors signify under- and overestimation of the gap, respectively. Given that tibial insert thickness was available in 1 mm increments, a gap estimation error of 1 mm was considered a clinically meaningful difference, since this error could affect the choice of insert thickness and subsequently alter ligament tension and balance.¹¹

Statistical analysis

Gap estimation errors were reported using means and SDs. Moreover, gap estimation errors > 1 mm were categorized in 1 mm increments and summarized using counts and proportions. To address our first research question, we reported the accuracy of gap estimation across all surgeons by pooling all surgeons and all cadavers. The proportion of visual gap assessments with a gap estimation error > 1 mm and > 2 mm was also calculated. Regarding our second research question, gap estimation errors were compared in extension versus flexion using an independent-samples t-test. In addition, gap estimation errors were compared across each of the four individual gaps that were measured (i.e., medial and lateral gaps in extension and medial and lateral gaps in flexion) using one-way analysis of variance (ANOVA) with Bonferroni correction for multiple comparisons (α = 0.05). Regarding our third research question, the gap estimation error pooled from all gap assessments for each surgeon (for both medial and lateral assessments in both extension and flexion) were compared by using the one-way ANOVA test with Bonferroni correction for multiple comparisons (α = 0.05). Regarding our fourth research question, the proportion of gap estimation errors exceeding the clinically significant threshold of 1 mm was compared by knee flexion angle and gap (i.e., medial and lateral gaps in extension and medial and lateral gaps in flexion) and, also, by surgeon using the chi-squared test with odds ratios and 95% CIs reported (α = 0.05). All analyses were two-tailed and performed using SPSS Statistics (v. 29.0, IBM, USA).

Results

Study sample and gap assessments

Six of the seven cadaveric knees were successfully assessed by all evaluators. In one specimen (cadaver six of seven), the medial collateral ligament (MCL) ruptured during assessment by the fourth surgeon, as indicated by an audible pop and a marked increase in valgus laxity. As a result, this specimen was excluded from further evaluation, and only the data from the first three surgeons’ assessments were retained for analysis.

In total, 78 gap assessments (39 medial, 39 lateral) were recorded in extension and 77 gap assessments (39 medial, 38 lateral) were recorded in flexion (Supplementary Material). One lateral gap measurement in flexion was not captured due to a technical error with the motion capture system during the chief resident’s evaluation of cadaver six; therefore, this specific gap estimation error was not calculated.

Gap accuracy of all surgeons

Regarding our first question, the mean gap estimation error across all surgeons and cadavers was -0.4 mm (SD 0.7) (Figure 2). The majority of gaps (111/155; 72%) were underestimated, while 35/155 (23%) were overestimated, and 9/155 (5%) were within the error bounds of the motion capture system (< 0.1 mm). Estimation error was within the calibration accuracy of the motion capture system (i.e., ≤ 0.1 mm) in 9/155 (6%) observations, > 1 mm in 48/155 (31%) observations, and > 2 mm in 9/155 (6%) observations.

Fig. 2.

Histograms describing distribution of gap estimation error across all cadavers and surgeons for a) all assessments and for each gap: b) medial extension, c) lateral extension, d) medial flexion, and e) lateral flexion. Positive error signifies overestimating the gaps (estimated > measured).

Gap accuracy by knee flexion angle and loading direction

Regarding our second question, the mean gap estimation error was greater in extension (-0.7 mm (SD 0.8)) compared to flexion (-0.2 mm (SD 1.0); p < 0.001, independent-samples t-test). Gaps were more frequently underestimated in extension (65/78; 83%) compared to flexion (46/77; 60%) (odds ratio (OR) 2.1 (95% CI 1.2 to 3.5); p = 0.002, independent-samples t-test). Moreover, differences in the gap estimation error as a function of gap and flexion angle were detected (p = 0.001, one-way ANOVA; Figure 3). Specifically, gap estimation error differed in lateral flexion compared to lateral and medial extension. The lowest mean estimation error was in lateral flexion (-0.1 mm (SD 1.0)), while the greatest was in lateral extension (-0.7 mm (SD 0.8)).

Fig. 3.

Distribution of gap estimation error for all surgeons by gap (positive error signifies overestimating the gaps (estimated > measured)). p-values were calculated using independent-samples t-test.

Gap accuracy by surgeon

Regarding our third question, overall gap estimation error differed between surgeons, ranging from a mean error of -0.8 mm (SD 0.8) to 0.2 mm (SD 1.2) (p < 0.001, one-way ANOVA) for all gap assessments. When comparing gap estimation errors by specific gap, surgeons differed in medial flexion (p = 0.003, one-way ANOVA) and lateral flexion (p = 0.016, one-way ANOVA) (Figure 4).

Fig. 4.

A box plot for medial and lateral gap estimation error in extension and flexion by surgeon, ordered based on years of experience where S1 (24 years), S2 (15 years), S3 (15 years), S4 (7 years), and S5 (5 years) are attending surgeons, and S6 is a chief resident. Brackets in red signify statistically significant differences between two surgeons (positive error signifies overestimating the gaps (estimated > measured)). p-values were calculated using one-way analysis of variance.

Clinically significant estimation errors

Regarding our fourth question, the proportion of clinically significant gap estimation errors (defined as > 1 mm) was 33% (26/78) in extension compared to 26% (20/77) in flexion corresponding to an OR of 1.4 (95% CI 0.7 to 2.8; p = 0.315, chi-squared test). At this error threshold for gap estimation, 22% (34/155) of all gap estimates were underestimated and 8% (12/155) were overestimated. When stratified by specific gap assessed, the proportion of clinically significant gap estimation errors was similar: 13/39 (33%) in lateral extension, 15/39 (39%) in medial extension, 10/38 (26%) in lateral flexion, and 10/39 (26%) in medial flexion (p = 0.566, chi-squared test). Finally, the proportion of clinically significant gap estimation errors by surgeon varied between 18% (5/28) to 42% (10/24) but these differences were not statistically significant (p = 0.370, chi-squared test). Similarly, no statistically significant differences were detected between surgeons when stratifying by specific gap.

Discussion

In this study of 155 gap assessments performed by six surgeons across seven cadaveric specimens, several key findings emerged. First, gap estimation errors greater than 1 mm occurred in approximately one-third of assessments, with 72% being underestimations. Second, both the magnitude and frequency of underestimation were greater in extension compared to flexion. Error also varied by the specific gap assessed, with the highest mean error observed in lateral extension and the lowest in lateral flexion. Third, estimation error varied among surgeons, particularly in medial and lateral flexion. However, no statistically significant differences were found in the frequency of clinically meaningful errors (> 1 mm) by flexion angle, gap location, or surgeon.

Achieving specific ranges of intraoperative gap measurements, which can be as narrow as 3 mm, correlate with improved pain outcomes on the Knee Injury and Osteoarthritis Outcome Score (KOOS).^9,10 Moreover, pain outcomes improve as more targeted gap ranges are met across medial and lateral compartments in flexion, midflexion, and extension.^9,10 In our study, we observed a gap estimation error greater than 1 mm in 31% (48/155) of all gap assessments, which represents one-third of the 3 mm gap range found to be critical for improved outcomes. We speculate that this finding supports the importance of accurate gap measurement to remain within the thresholds that are linked to greater patient satisfaction following TKA.

While the mean gap estimation error was statistically greater in extension (−0.7 mm) than in flexion (−0.2 mm), the clinical importance of the magnitude of the mean errors is debatable. We defined a clinically significant error as an estimation exceeding 1 mm, which is a magnitude that can impact final insert thickness and soft-tissue tension. We found that 13 of the 39 medial extension gap assessments (33%) were underestimated by ≥ 1 mm. This is notable considering prior studies that highlight the critical role of medial compartment stability in successful outcomes and the relationship between MCL tension and AP laxity in PS-TKA.^12,15 For instance, assuming a population mean MCL stiffness of 80 N/mm, a 1 mm underestimate in the medial gap could decrease the resting tension of the MCL by approximately 80 N.¹⁵ Berube et al¹² reported that reduced MCL tension can lead to clinically significant increases in anterior-posterior laxity exceeding 1 cm.

Our group of six surgeons underestimated gaps about one-third of the time, more often in extension than in flexion (Figure 3). A possible reason for underestimating the medial gap in extension is limited visualization: the anterior lip of the polyethylene insert may block the view of the femoral-tibial contact point. This challenge may be greater with more congruent liners, such as medial congruent designs, which have a larger anterior lip. For the lateral extension gap, partial obstruction by the subluxated patella may also contribute to estimation errors. In flexion, posterior translation of the femoral component in PS TKA improves visibility of the femoral-tibial contact point and resultant gap, and could help explain more accurate gap estimates at this angle.

This study has limitations. First, repeated testing on each cadaveric leg could have stretched the collateral ligaments. While the order of surgeons was randomized to mitigate potential systematic changes in ligamentous laxity from repeated testing, some alteration in tissue properties over the course of the experiment cannot be entirely excluded. Further, cadaveric soft-tissues do not fully replicate the dynamic, physiological behaviour of in vivo tissues in TKA, which are influenced by muscle tone, anaesthesia, and blood flow.¹⁶ Despite these constraints, the cadaveric model was essential for this study’s design as it provided a stable and controlled environment, allowing for repeated assessment of knee balance by multiple surgeons. To mitigate the effect of repeated examination, the first surgeon examiner repeated his assessment at the conclusion of testing and confirmed that his choice of insert thickness had not changed. Second, our findings in a posterior stabilized implant system cannot be generalized to implant systems that preserve one or both cruciate ligaments or more conforming bearing surfaces. Such medial congruent articulations, with a taller anterior lip, could further increase the difficulty of accurately assessing the medial gap in extension. Nevertheless, condylar, posterior cruciate ligament-substituting TKA systems with relatively non-conforming bearing surfaces comprise a large percentage of TKA systems used in the USA, which supports the clinical applicability of our findings.¹⁷ Third, a limitation of this study is the relatively small number of participating surgeons. While we detected statistically significant differences in gap estimation accuracy between surgeons, the magnitude of error was not clinically meaningful (< 1 mm). This study was not designed to evaluate individual surgeon performance or the factors contributing to their specific error patterns. Therefore, these inter-surgeon findings should be interpreted with caution. Fourth, since this study included surgeons who were trained at or practice at the same institution, the findings may not be generalizable to all surgeons; however, the results are likely representative of surgeons working at high-volume academic medical centres. Fifth, the small sample size of seven restricts the generalizability of our findings, as the human cadavers we used may not represent the diverse patient populations. Nevertheless, we used appropriate statistical analysis to analyze the collected data and derive our conclusions. Future work will focus on expanding our assessment across a larger population, either in vitro or in vivo, to increase the generalizability of our findings. Sixth, the cadaveric design of this study allowed for precise, controlled measurement of gap errors, but it precludes any assessment of clinical outcomes. Therefore, we were unable to draw a direct causal line between the estimation errors we quantified and in vivo consequences such as joint instability, pain, or patient dissatisfaction. The clinical relevance of our findings is therefore inferred by connecting the magnitude of the observed errors to established principles of knee balancing and the extensive clinical literature linking imbalance to adverse outcomes. Finally, the measurement error of the motion capture system (0.1 mm) was acceptable, given that this is about ten-fold greater than the associated increase or decrease in gap measurement associated with selecting a 1 mm thicker or thinner tibial insert.

In conclusion, within the constraints of this single-centre study involving a small cohort of surgeons and a single implant system, surgeon error in visual assessment of medial and lateral gaps in extension and flexion exceeded 1 mm up to one-third of the time. Gap underestimation, especially in extension, was the most common error among surgeons. This finding reveals an opportunity for improvement in this critical aspect of TKA once findings are substantiated in a larger, multicentre cohort.

Take home message

- The authors found that, in posteriorly stabilized total knee arthroplasty, surgeons tend to underestimate intraoperative gaps during knee balance exam, particularly in extension.

- The authors also found that clinically meaningful gap errors (> 1 mm) occurred in up to 33% of exams, which suggests a need to enhance the current manual gap assessment method.

Data Availability

All data generated or analyzed during this study are included in the published article and/or in the supplementary material.