Acetabular component placement accuracy and short-term outcomes in total hip arthroplasty

Abstract

Aims

Precise acetabular component positioning is critical to the success of total hip arthroplasty (THA). Mako uses CT-based robotic assistance, whereas AR-hip is an augmented reality-based portable navigation system operating without preoperative CT and at lower costs. Although both systems improve accuracy, direct comparative data are scarce. We compared cup placement accuracy and short-term outcomes between Mako robotic arm and AR navigation systems.

Methods

This single-centre retrospective cohort study included 192 primary THAs (Mako, 147; AR navigation, 45). Cup placement accuracy was assessed three-dimensionally using postoperative CT, evaluating cup alignment (radiological inclination (RI) and anteversion (RA)) and position (centre of rotation (COR)). Short-term outcomes were evaluated using the Harris Hip Score (HHS), Japanese Orthopaedic Association hip score (JOA), Japanese Hip-disease Evaluation Questionnaire (JHEQ), Timed Up and Go (TUG), and 10 m walk test (10MWT). Baseline differences were adjusted using propensity score overlap weighting.

Results

Mako demonstrated significantly superior accuracy than AR navigation, with smaller absolute errors in cup angles (RI 2.0° vs 3.0°, p = 0.045; RA 1.7° vs 2.4°, p = 0.005) and COR errors in the superior–inferior direction (1.5 mm vs 3.7 mm, p < 0.001). More cases were within 5° of planned angles with Mako (92.3% vs 77.7%, p = 0.030), while Lewinnek safe zone rates were high in both groups (99.5% vs 95.2%, p = 0.202). Short-term clinical scores and functional recovery at discharge were similar between groups. Operating time and blood loss were greater with Mako (122.6 vs 103.5 mins; 242.5 vs 171.5 ml; both p < 0.001). Subgroup analyses showed greater advantages of Mako in complex cases, such as obesity and developmental dysplasia of the hip, although interactions were insignificant.

Conclusion

Mako demonstrated superior cup placement accuracy and COR reconstruction compared with AR navigation, with comparable short-term clinical outcomes, potentially offering advantages in complex cases. However, further research is warranted to clarify its long-term outcomes and cost-effectiveness.

Cite this article: Bone Jt Open 2026;7(2):203–215.

Introduction

Total hip arthroplasty (THA) is a well-established treatment for end-stage hip disease that improves quality of life.¹ Precise positioning of the acetabular component is crucial, as malpositioning can cause impingement, dislocation, polyethylene wear, aseptic loosening, and premature revision, ultimately compromising implant longevity.²

Computer-assisted surgery (CAS) technologies were developed to improve implant placement.³ Robotic arm-assisted THA systems, such as the Mako system (Stryker, USA), integrate CT-based preoperative planning with real-time intraoperative robotic guidance for precise and reproducible cup positioning.^4,5 Clinical studies report improved accuracy, reduced postoperative pain, and accelerated functional recovery.⁶

More recently, portable navigation systems based on augmented reality (AR) have gained significant attention. The AR Hip-navigation system (Zimmer Biomet Japan, Japan) utilizes smartphone technology to overlay planned cup orientation onto the surgical field in real time. By leveraging the device’s gyroscope to detect gravitational vectors, this system provides surgical guidance without requiring preoperative CT imaging, offering advantages in simplicity, cost-effectiveness, and accessibility, while showing encouraging preliminary results compared with conventional techniques.^7,8

Despite growing adoption, direct comparative studies between robotic and AR navigation systems remain scarce. Only one study has assessed Mako, AR, and CT-based navigation simultaneously,⁹ with no reports directly comparing Mako with AR Hip systems. A recent network meta-analysis confirmed this gap.¹⁰ This study aimed to directly compare acetabular component placement accuracy and short-term clinical outcomes between Mako robotic arm assistance and AR navigation in primary THA, to clarify their respective advantages and limitations. We hypothesized that the Mako system would achieve higher cup placement accuracy and better short-term outcomes than the AR navigation system.

Methods

This single-centre retrospective cohort study was approved by the Institutional Review Board and conducted per the Declaration of Helsinki. The STROBE guidelines were followed.¹¹

We included patients who underwent primary THA using either robotic arm assistance (Mako) or AR navigation (AR Hip) between October 2021 and June 2024. Eligible diagnoses included osteoarthritis (OA), osteonecrosis of the femoral head (ONFH), and rheumatoid arthritis (RA). Inclusion required intraoperative cup data and postoperative CT imaging. Exclusion criteria were revision surgery, trauma, and active infection.

System use followed institutional timelines: AR navigation (October 2021 to May 2023), followed by the progressive adoption of Mako from July 2022. Since June 2023, all THAs have been Mako-assisted.

Participants and baseline characteristics

Of 245 patients screened for primary THA, 53 were excluded (49, missing data; four, femoral neck fractures), leaving 192 procedures (Mako, 147; AR navigation, 45) for analysis (Figure 1). After overlap weighting, baseline covariates were balanced, with all weighted SMDs < 0.1 (Figure 2, Table I).

Fig. 1.

Flowchart of patient inclusion. AR, augmented reality.

Fig. 2.

Covariate balance before and after overlap weighting (standardized mean difference ≤ 0.1 indicates acceptable balance).

Table I.

Baseline characteristics of patients undergoing total hip arthroplasty using Mako compared with augmented reality (AR) navigation systems before and after propensity score overlap weighting.

Characteristic	Before overlap weighting		SMD	After overlap weighting		SMD
	Mako	AR navigation		Mako	AR navigation
Patients, n	147	45
Age, yrs	71.0 (64.0 to 76.0)	65.0 (59.0 to 72.0)	0.496	66.3 (9.7)	66.3 (8.1)	< 0.001
Sex, n (%)			0.076			< 0.001
Female	127 (86)	40 (89)		27.8 (86.9)	27.8 (86.9)
Male	20 (14)	5 (11)		4.2 (13.1)	4.2 (13.1)
Laterality, n (%)			0.202			< 0.001
Right	93 (63)	24 (53)		17.8 (55.6)	17.8 (55.6)
Left	54 (37)	21 (47)		14.2 (44.4)	14.2 (44.4)
BMI, kg/m²	23.7 (21.4 to 26.8)	24.5 (23.3 to 28.3)	0.261	25.2 (0.5)	25.2 (0.7)	< 0.001
Diagnosis, n (%)			0.301			< 0.001
Primary OA	20 (14)	5 (11)		3.9 (12.3)	3.9 (12.3)
DDH
Crowe I	89 (61)	27 (60)		19.8 (61.9)	19.8 (61.9)
Crowe II	18 (12)	5 (11)		3.4 (10.7)	3.4 (10.7)
Crowe III	2 (1.4)	1 (2.2)		0.5 (1.7)	0.5 (1.7)
RA	5 (3.4)	4 (8.9)		2.0 (6.1)	2.0 (6.1)
ONFH	8 (5.4)	2 (4.4)		1.7 (5.3)	1.7 (5.3)
RDC	3 (2.0)	1 (2.2)		0.7 (2.0)	0.7 (2.0)
Post-traumatic OA	2 (1.4)	0 (0)		0 (0)	0 (0)
Surgical approach, n (%)			0.129			< 0.001
MWJ	146 (99)	44 (98)		31.5 (98.5)	31.5 (98.5)
DLA	1 (0.7)	1 (2.2)		0.5 (1.5)	0.5 (1.5)
Pelvic tilt angle, °	4.3 (-0.8 to 8.8)	4.1 (1.7 to 7.6)	0.170	4.4 (6.8)	4.4 (5.1)	< 0.001

Continuous variables are presented as median (IQR) before weighting and as mean (SD) after overlap weighting.

DDH, developmental dysplasia of the hip; DLA, direct lateral approach; MWJ, Modified Watson-Jones approach; OA, osteoarthritis; ONFH, osteonecrosis of the femoral head; RA, rheumatoid arthritis; RDC, rapidly destructive coxarthrosis; SMD, standardized mean difference.

Preoperative planning

CT scans (1.0 mm slice, pelvis to distal femur) were processed using ZedHip (LEXI, Japan) to generate 3D models referenced to the functional pelvic plane (FPP). ZedHip is a CT-based 3D preoperative planning and measurement software commonly used for 3D planning in THA.^12,13 The target cup orientation was a radiological inclination (RI) of 40° and radiological anteversion (RA) of 15°. The hip centre of the cup was planned to reproduce the anatomical acetabular centre, corresponding to the native centre of rotation (COR), using the contralateral hip as a reference. For centre-edge angle (CEA) < 0°, the cup was positioned superiorly to achieve CEA ≥ 0°, ensuring adequate superolateral coverage.¹⁴ Minor adjustments were made when bone deficiency or dysplasia required optimization of coverage and fixation (Supplementary Material).

Surgical technique

All procedures were performed under general anaesthesia by a single experienced hip surgeon (TN). At the beginning of this study, the surgeon had limited prior experience with robotic arm-assisted THA and had only recently started using the AR navigation system. However, the surgeon had substantial experience with accelerometer-based portable navigation systems, which share a similar operating concept. Patients were placed in the lateral decubitus position, using the modified Watson-Jones approach as the standard. A direct lateral approach was used for severe deformities or restricted motion. Cementless acetabular components with optional screw fixation were used in all cases. Cementless femoral components were the standard, with cemented components for poor-quality cases. Details of the implant systems used in each group are presented in the Supplementary Material. Full weightbearing was permitted from postoperative day 1.

Mako used CT-based registration, replicating ZedHip planning with identical CT data and coordinate systems; anterior superior iliac spine (ASIS) alignment ensured consistency. The robotic arm provided haptic control of reaming and cup placement. The AR system defined the FPP from ASIS and gravity and provided angle-only guidance during implantation. Intraoperative RI and RA were recorded in both systems for error analysis. Detailed workflows have been described previously,^4,7,9 and are summarized in the Supplementary Material.

Postoperative CT evaluation

Cup placement accuracy was evaluated using ZedHip based on postoperative CT scans acquired at the first outpatient follow-up. The postoperative implant model was automatically aligned with the FPP defined during preoperative planning, allowing comparison with the plan. Placement errors were defined as differences between planned and postoperative values for RI, RA, and centre of rotation (COR).

The coordinate system was defined as follows: the x-axis (medial-lateral, ML) connecting the bilateral ASIS, z-axis (superior-inferior, SI) perpendicular to the x-axis within the FPP, and y-axis (anterior-posterior, AP) perpendicular to both. Angular errors (RI and RA) were measured in degrees, and COR errors were calculated as absolute deviations in the ML, AP, and SI directions (mm).

ZedHip has demonstrated high reliability, with intraclass correlation coefficients (ICCs) > 0.9 for both inter-rater and intrarater assessments,^15,16 and our study followed these validated protocols.

Variables

Baseline variables included age, sex, height, weight, BMI, diagnostic category (primary OA; developmental dysplasia of the hip (DDH)-related OA, including post-Chiari osteotomy and Crowe types I to IV;¹⁷ RA; ONFH; rapidly destructive coxarthrosis (RDC); and post-traumatic OA), laterality, surgical approach, pelvic tilt angle (defined as the angle between the anterior pelvic plane and the FPP, measured on preoperative CT),¹⁸ and the number of days between surgery and postoperative CT.

Outcomes

The primary outcome was cup placement accuracy assessed through angular and positional errors. Angular errors were calculated as absolute differences (°) in 1) planned versus postoperative values (accuracy error); 2) planned versus intraoperative values (operational errors); and 3) intraoperative versus postoperative values (navigation errors).

The first and third definition followed previous studies,¹⁹ whereas the second was newly defined to assess intraoperative execution accuracy. Accuracy error was considered the most clinically relevant.

The positional accuracy was the absolute COR deviation (mm) from the plan in the ML, AP, and SI directions. We calculated proportions within Lewinnek’s safe zone (RI 40° (± 10°), RA 15° (± 10°))²⁰ and stricter within 5° thresholds per prior studies.^18,19,21 We also calculated the proportion within 3 mm and 5 mm, as previous reports deemed these thresholds acceptable for ML and SI.^5,21,22

Secondary outcomes included perioperative factors (operating time, blood loss, and in-hospital complications, e.g. dislocation, infection, reoperation, and fracture).

Clinical outcomes were assessed using the Harris Hip Score (HHS)^23,24 and the Japanese Orthopaedic Association (JOA) hip score.²⁵ Patient-reported outcomes were evaluated using the Japanese Hip-joint Evaluation Questionnaire (JHEQ).²⁶ Physical function was assessed using the Timed Up and Go (TUG) test^27,28 and the 10 m walk test (10MWT).^28,29

All clinical and functional metrics were measured preoperatively and at discharge. Further details of scoring systems and test protocols are provided in the Supplementary Material.

Statistical analysis

Categorical variables were presented as counts and percentages, and continuous variables as medians with IQR. Categorical variables were compared using the chi-squared test or Fisher’s exact test, as appropriate, and continuous variables were compared using the Mann-Whitney U test. Complete case analysis was performed for primary outcomes. For secondary outcomes, missing data were handled by multiple imputation as a sensitivity analysis.

Overlap weighting (average treatment effect in the overlap population; ATO), based on propensity scores, was used to adjust for potential confounders. This method assigns greater weight to individuals with overlapping covariate distributions across groups, reducing extrapolation and improving estimation stability.³⁰ Propensity scores were estimated by logistic regression using the following covariates: age, sex, BMI, diagnosis, laterality, surgical approach, and pelvic tilt angle. Overlap weighting (ATO) was implemented using the WeightIt package in R (v4.5.0, method = ‘ps’, estimand = ‘ATO’). Covariate balance was assessed using standardized mean differences (SMDs), with SMD < 0.1 indicating adequate balance,³¹ calculated using the cobalt package.

Since this was a retrospective observational study, no a priori sample size calculation was performed. Instead, all eligible cases during the study period were included to maximize representativeness. To evaluate the adequacy of the sample size, post hoc statistical power was estimated based on the effective sample size, accounting for the reduction due to overlap weighting.³² The power calculations were performed using the observed between-group differences and overlap-weighted SDs of the primary outcomes.

The treatment effects were estimated using overlap-weighted linear or logistic regression models. Analyses were performed using the svyglm function of the survey package in R, and results were reported as weighted means or odds ratios with robust standard errors (SEs) and 95% CIs. For score-based secondary outcomes, change-from-baseline (Δ) values were used as dependent variables, and baseline scores were additionally included as covariates to adjust for pre-treatment imbalance.³³

Prespecified subgroup analyses examined outcome heterogeneity by BMI (< 30 kg/m² vs ≥ 30 kg/m²), pelvic tilt angle (< 0° vs ≥ 0°), and Crowe classification (Type I vs ≥ Type II). Interaction p-values were derived from weighted regression models including interaction terms.

Sensitivity analyses included adjustment for the number of days from surgery to postoperative CT, given the potential influence on cup position and angle measurements. For secondary outcomes, multiple imputation using the mice package in R was performed, and analyses without baseline score adjustment were also conducted to assess this covariate’s influence.³⁴

Statistical significance was set at p < 0.05, with all analyses performed using R software (v4.5.0; R Foundation for Statistical Computing, Austria).

Results

Primary outcome

Regarding cup angle, accuracy errors were significantly smaller with Mako for RI (2.0° (SE 0.2°) vs 3.0° (SE 0.4°); −0.9°, 95% CI −1.9 to 0.0; p = 0.045) and RA (1.7° (SE 0.2°) vs 2.4° (SE 0.2°); −0.8°, 95% CI −1.3 to −0.3; p = 0.005) (Figure 3, Table II). Navigation errors in RI were not significantly different, whereas RA navigation error was significantly smaller with Mako. Operational errors were significantly smaller with Mako for both RI and RA.

Fig. 3.

Weighted distribution of cup alignment and positioning errors in Mako and augmented reality (AR) navigation systems. a) Distribution of radiological inclination (RI) and anteversion (RA). b) Distribution of cup position errors in the medial-lateral (ML) and superior-inferior (SI) directions. Positive ML = lateral, negative = medial; positive SI = superior, negative = inferior. Point size represents normalized overlap weights (range, 0.0 to 1.0).

Table II.

Comparison of cup placement angles and positional accuracy between Mako and augmented reality (AR) navigation systems after overlap weighting.

Variable	Mako	AR navigation	Difference (95% CI)	p-value
Cup angle accuracy (RI and RA)
Cup placement angles measured postoperatively
RI, °	38.3 (0.2)	42.0 (0.6)	-3.7 (-4.9 to –2.6)	< 0.001
RA, °	14.7 (0.2)	15.3 (0.4)	-0.6 (-1.5 to 0.4)	0.234
Cup placement angles measured intraoperatively
RI, °	39.8 (0.1)	40.6 (0.6)	-0.7 (-1.9 to 0.5)	0.249
RA, °	15.0 (0.1)	15.8 (0.3)	-0.8 (-1.4 to –0.2)	0.006
Absolute value of accuracy error
RI, °	2.0 (0.2)	3.0 (0.4)	-0.9 (-1.9 to 0.0)	0.045
RA, °	1.7 (0.2)	2.4 (0.2)	-0.8 (-1.9 to 0.0)	0.005
Absolute value of navigation error
RI, °	2.2 (0.2)	2.3 (0.3)	-0.2 (-0.8 to 0.5)	0.599
RA, °	1.6 (0.1)	2.3 (0.3)	-0.7 (-1.4 to –0.1)	0.04
Absolute value of operational error
RI, °	0.7 (0.1)	2.7 (0.5)	-2.0 (-2.9 to –1.1)	< 0.001
RA, °	0.7 (0.1)	1.5 (0.2)	-0.9 (-1.3 to –0.5)	< 0.001
Within Lewinnek’s safe zone, %	99.5 (0.5)	95.2 (3.3)	4.3 (-2.3 to 10.9)	0.204
RI (30° to 50°), %	100.0 (0.0)	95.2 (3.3)	4.8 (-1.7 to 11.4)	0.151
RA (5° to 25°), %	99.5 (0.5)	100.0 (0.0)	-0.5 (-1.5 to 0.5)	0.319
Within 5° of planned angle, %	92.3 (2.2)	77.7 (6.3)	14.6 (1.4 to 27.7)	0.031
RI (35° to 45°), %	96.5 (1.3)	82.4 (5.8)	14.2 (2.6 to 25.8)	0.017
RA (10° to 20°), %	94.9 (1.9)	95.4 (3.2)	-0.5 (-7.8 to 6.9)	0.903
Cup position accuracy (ML, AP, SI)
Cup position error, mm
ML error	1.5 (0.1)	2.0 (0.2)	-0.5 (-1.0 to 0.0)	0.065
AP error	2.0 (0.2)	1.7 (0.1)	0.3 (-0.1 to 0.7)	0.196
SI error	1.5 (0.10)	3.7 (0.3)	-2.2 (-2.8 to –1.6)	< 0.001
Positional accuracy (within 5 mm)
ML and SI, %	98.9 (0.7)	72.4 (6.7)	26.5 (13.2 to 39.8)	< 0.001
ML, %	99.7 (0.3)	93.4 (3.8)	6.3 (-1.2 to 13.8)	0.101
SI, %	99.2 (0.6)	79.0 (6.1)	20.2 (8.2 to 32.2)	0.001
Positional accuracy (within 3 mm)
ML and SI, %	77.3 (3.8)	22.5 (6.4)	54.8 (40.2 to 69.3)	< 0.001
ML, %	89.3 (3.1)	69.7 (7.1)	19.6 (4.5 to 34.7)	0.012
SI, %	88.0 (2.7)	29.9 (7.0)	58.1 (43.4 to 72.8)	< 0.001

All values are expressed as mean (standard error) after overlap weighting.

Difference is calculated as Mako – AR navigation. All p-values were obtained using weighted regression models.

AP, anteroposterior; AR, augmented reality; ML, medial-lateral; RA, radiological anteversion; RI, radiological inclination; SI, superior-inferior.

The proportion within 5° of the plans was significantly higher with Mako (92.3% (2.2%) vs 77.7% (6.3%); risk difference 14.6%, 95% CI 1.4 to 27.7; p = 0.030). Lewinnek safe zone attainment was high in both groups. Odds ratios (ORs) from weighted logistic regression for binary outcomes with near-complete success are provided in the Supplementary Material.

Cup position (COR) accuracy

Positional accuracy strongly favoured Mako, particularly in the SI directions (1.5 mm vs 3.7 mm, p < 0.001), while ML and AP differences were not statistically significant (Figure 3, Table II). Mako achieved significantly higher rates of acceptable positioning within 5 mm in both ML and SI directions and within the stricter margin of 3 mm.

Secondary outcomes

Operating time was significantly longer (122.6 mins (SE 1.6) vs 103.5 mins (SE 1.8); p < 0.001) and BL significantly higher (242.5 ml (SE 12.4) vs 171.5 ml (SE 15.7); p < 0.001) with Mako (Table III and Table IV).

Table III.

Comparison of secondary outcomes between Mako and augmented reality (AR) navigation systems after overlap weighting, adjusted between-group comparisons after baseline score adjustment.

Outcomes	Mako	AR navigation	Difference (95% CI)	p-value
Operating time (min)	122.6 (1.6)	103.5 (1.8)	19.1 (14.3 to 23.9)	< 0.001
Blood loss (ml)	242.5 (12.4)	171.5 (15.7)	71.0 (31.5 to 110.5)	< 0.001
In-hospital surgical complications	0	0
ΔJOA score	25.4 (1.1)	25.5 (2.0)	-0.1 (-4.6 to 4.4)	0.953
ΔHarris Hip Score	19.2 (1.2)	18.5 (2.8)	0.7 (-5.2 to 6.6)	0.814
ΔJHEQ total	27.0 (1.3)	23.9 (2.6)	3.0 (-2.7 to 8.8)	0.300
ΔVAS	-57.9 (2.3)	-59.8 (2.7)	1.9 (-5.0 to 8.9)	0.577
ΔPain	12.5 (0.9)	10.7 (0.9)	1.8 (-0.2 to 3.8)	0.081
ΔMovement	7.4 (0.5)	7.8 (0.9)	-0.4 (-2.5 to 1.7)	0.728
ΔMental	7.1 (0.5)	6.6 (0.9)	0.5 (-1.5 to 2.6)	0.618
ΔTUG (sec)	-3.2 (0.2)	-3.6 (0.3)	0.4 (-0.1 to 1.0)	0.129
Δ10MWT (m/s)	0.18 (0.02)	0.21 (0.03)	-0.03 (-0.11 to 0.04)	0.404

Adjusted between-group differences were calculated using analysis of covariance after overlap weighting, controlling for baseline values. Δ indicates postoperative minus preoperative change (unadjusted). Values are presented as weighted means (standard errors). Between-group differences are presented with 95% CIs and p-values derived from weighted linear regression models.

JHEQ, Japanese Hip-joint Evaluation Questionnaire; JOA, Japanese Orthopaedic Association; 10MWT, 10-Metre Walk Test; TUG, Timed Up and Go test; VAS, visual analogue scale.

Table IV.

Comparison of secondary outcomes between Mako and augmented reality (AR) navigation systems after overlap weighting, unadjusted pre- and postoperative values used for the analysis.

Outcomes	Preoperative	Postoperative	Δ(Post-Pre)
JOA score
Mako	45.7 (1.4)	72.0 (1.2)	26.3 (1.6)
AR navigation	50.3 (2.7)	73.3 (2.1)	23.0 (2.7)
Harris Hip Score
Mako	56.2 (1.7)	75.2 (1.1)	19.1 (1.9)
AR navigation	59.7 (2.5)	75.4 (3.0)	15.8 (3.0)
JHEQ total
Mako	23.6 (1.3)	51.4 (1.4)	27.8 (1.5)
AR navigation	26.4 (2.5)	49.6 (2.9)	23.2 (2.8)
VAS
Mako	76.9 (2.5)	18.7 (2.3)	-58.4 (3.1)
AR navigation	74.1 (4.6)	16.4 (2.7)	-57.7 (4.9)
Pain
Mako	8.8 (0.6)	21.8 (0.5)	13.0 (0.7)
AR navigation	8.7 (1.0)	20.0 (0.9)	11.3 (1.2)
Movement
Mako	5.7 (0.5)	13.3 (0.5)	7.5 (0.6)
AR navigation	7.0 (1.0)	14.1 (1.0)	7.1 (1.1)
Mental
Mako	9.6 (0.5)	16.7 (1.2)	7.2 (0.5)
AR navigation	10.1 (1.0)	16.5 (1.2)	6.5 (1.0)
TUG (sec)
Mako	13.1 (0.8)	9.9 (0.3)	-3.1 (0.7)
AR navigation	10.9 (0.7)	8.8 (0.3)	-2.1 (0.5)
10MWT (m/s)
Mako	1.24 (0.04)	1.40 (0.03)	0.16 (0.03)
AR navigation	1.18 (0.05)	1.40 (0.04)	0.22 (0.04)

Unadjusted pre- and postoperative values (raw data) used for analysis of covariance are presented for reference.

JHEQ, Japanese Hip-joint Evaluation Questionnaire; JOA, Japanese Orthopaedic Association; 10MWT, 10-Metre Walk Test; TUG, Timed Up and Go test; VAS, visual analogue scale.

No significant differences were observed between groups in terms of clinical scores or functional outcomes. Baseline characteristics of complete-case datasets used in each secondary outcome analysis are provided in the Supplementary Material and demonstrate adequate covariate balance after overlap weighting (all SMDs < 0.1).

Subgroup analyses

Mako’s advantage in accuracy was directionally consistent across BMI (< 30 kg/m² vs ≥ 30 kg/m²), pelvic tilt (< 0° vs ≥ 0°), and Crowe classification subgroups (Type I vs ≥ Type II), though interaction p-values were not statistically significant (Table V). Point estimates suggested greater benefits in patients with BMI ≥ 30 kg/m² for cup angle accuracy and in Crowe II to III cases for cup position accuracy, but CIs were wide due to limited sample sizes (Figure 4).

Table V.

Subgroup analysis of cup placement accuracy between Mako and augmented reality (AR) navigation systems.

Variable	Mako (%)	AR navigation (%)	Risk difference, % (95% CI)	p-value	Interaction p-value
Cup angle accuracy (RI and RA < 5°)
Overall	92.3 (2.2)	77.7 (6.3)	14.6 (1.4 to 27.7)	0.031	-
BMI, kg/m 2					0.507
< 30	93.9 (1.9)	81.9 (6.7)	12.0 (-1.6 to 25.6)	0.086
≥ 30	84.4 (8.8)	59.2 (16.4)	25.2 (-11.9 to 62.4)	0.196
Pelvic tilt angle, °					0.706
＜ 0	96.1 (2.3)	86.4 (12.6)	9.7 (-15.7 to 35.1)	0.457
≥ 0	91.1 (2.8)	75.8 (7.2)	15.4 (0.3 to 30.4)	0.048
DDH Crowe classification					0.770
I	90.0 (3.0)	77.9 (8.1)	12.1 (-4.8 to 29.1)	0.164
II to III	87.5 (9.0)	68.7 (78.9)	18.7 (-22.9 to 60.4)	0.387
Cup position accuracy (ML and SI errors < 5 mm)
Overall	98.9 (0.7)	72.4 (6.7)	26.5 (13.2 to 39.8)	< 0.001	-
BMI, kg/m 2					0.393
< 30	98.6 (0.8)	69.9 (7.8)	28.7 (13.3 to 44.0)	< 0.001
≥ 30	100 (0.0)	83.1 (11.3)	16.9 (-5.7 to 39.5)	0.156
Pelvic tilt angle, °					0.774
＜ 0	96.5 (2.5)	65.5 (16.8)	31.0 (-2.5 to 64.5)	0.077
≥ 0	99.6 (0.4)	73.9 (7.3)	25.7 (11.3 to 40.1)	< 0.001
DDH Crowe classification					0.382
I	98.7 (1.0)	82.8 (7.2)	15.9 (1.7 to 30.1)	0.03
II to III	100 (0.0)	65.4 (20.1)	34.6 (-5.3 to 74.6)	0.102

Interaction p-values represent the test for heterogeneity across subgroups.

Wide CIs in certain subgroups (e.g. Crowe II to III, BMI ≥ 30 kg/m²) may reflect limited sample sizes should be interpreted with caution.

All values are presented as weighted proportions (standard error) after overlap weighting.

DDH, developmental displaysia of the hip; ML, medial-lateral; RA, radiological anteversion; RI, radiological inclination; SI, superior-inferior.

Fig. 4.

Forest plot of subgroup analyses for cup placement accuracy between Mako and augmented reality (AR) navigation systems. a) Risk difference for achieving accurate cup alignment (radiological inclination and radiological anteversion < 5°). b) Risk difference for achieving accurate cup position (medial-lateral and superior-inferior errors < 5 mm). Positive values favour Mako. DDH, developmental displaysia of the hip.

Sensitivity analyses

Additional adjustments for surgery-to-CT interval did not change the primary outcome estimates. Multiple imputation analyses for secondary outcomes yielded results consistent with complete-case analyses, and models without baseline score adjustment showed similar effect directions.

In the post hoc power analysis, the total sample size was 192, and the variance inflation factor (VIF) due to weighting was 1.50, yielding an effective sample size of approximately 128. Using overlap-weighted SDs for cup angle errors (RI: 2.33°, RA: 1.60°) and a two-sided α = 0.05, the estimated statistical powers to detect the observed between-group differences (ΔRA ≈ 0.8°, ΔRI ≈ 0.9°) were approximately 0.80 and 0.60, respectively. These results indicate adequate statistical power for detecting differences in RA and moderate power for RI, supporting the stability and representativeness of the weighted analyses.

Discussion

In this propensity score-weighted analysis, Mako robotic arm-assisted THA demonstrated superior acetabular component placement accuracy compared with AR navigation, while short-term clinical outcomes remained comparable between groups. While several studies have confirmed the high accuracy of robotic-assisted THA, few have directly compared robotic and AR-based navigation under an identical 3D-CT framework. This direct head-to-head analysis therefore provides complementary evidence on the relative precision and reproducibility of these two computer-assisted systems.

Our findings confirm Mako’s superior angular precision, consistent with previous comparative studies.⁹ In this study, Mako showed reduced errors across all phases, with particular superiority in intraoperative reproducibility and final placement accuracy. CT navigation offers higher intraoperative accuracy than accelerometer systems, but this often does not improve final placement accuracy.¹⁹ By separating accuracy errors into navigation and operational components, the present study clarified that the robotic arm’s haptic control mainly reduces operation-related variability rather than navigation accuracy itself. Mako not only ensures high navigation accuracy but also provides superior stability during reaming and impaction, resulting in minimal errors throughout all phases.⁴ These findings suggest that the robotic arm’s haptic boundaries and guided workflow effectively suppress intraoperative deviations and enable superior final accuracy.

Regarding positional accuracy, the Mako group exhibited smaller errors in all directions than the AR group, with a statistically significant difference in the SI direction. Balanced hip reconstruction requires COR restoration within 5 mm horizontally and 3 mm vertically of the anatomical position.²² These thresholds were achieved more consistently with Mako. 2D scatter plots visualizing the directional distribution of COR deviations clearly demonstrated this trend, showing that the AR group more frequently exhibited lateral and inferior deviations. These deviations likely reflecting under-reaming or incomplete cup seating. Robotic arm assistance ensures consistent reaming depth and sphericity, enabling reliable press-fit fixation. This precision increases bone-cup contact and frictional resistance, enhancing rotational stability and lever-out strength.³⁵ This secure initial fixation promotes osseointegration and may lower the long-term risk of aseptic loosening.

The advantage of robotic assistance was numerically greater in patients with obesity and those with Crowe II to III DDH, with higher proportions achieving the target of within 5°, consistent with previous reports showing improved reproducibility in such technically demanding cases.^36,37 Similar trends were observed for cup position accuracy, and the superiority of Mako was maintained across pelvic tilt subgroups, even in the posterior pelvic tilt, which has been reported as a risk factor for cup angle errors.¹⁸ However, none of the interaction tests reached statistical significance, and CIs were wide owing to limited sample size. Taken together, these findings suggest that robotic assistance may be particularly beneficial in challenging cases, although interpretation requires caution.

When the difference in cup placement accuracy observed in this study is interpreted from a clinical perspective, the mean angular error between the two groups was small (approximately 1°) and likely not clinically meaningful. The Mako group showed a higher proportion of cups within 5° of the planned values, indicating greater reproducibility. This was particularly evident in technically demanding cases, such as in obese patients or those with Crowe type II to III deformities. The concept of patient-specific safe zones, which account for individual spinopelvic mobility and alignment, has recently been proposed, and deviations greater than 5° to 6° in version or 4° in inclination have been linked to postoperative instability.^38,39 The 14.6% higher rate of cups within 5° in the Mako group underscores the clinical importance of accurately reproducing the planned cup orientation. In terms of positional accuracy, the Mako group exhibited smaller mean errors, particularly in the superior-inferior direction, with a mean difference of approximately 2 mm. The higher proportion of cups positioned within 3 mm and 5 mm of the planned target was also significantly higher, demonstrating superior reproducibility. As even a 3 mm vertical deviation may alter the load axis and hip biomechanics,²² the 2 mm difference observed in this study could be clinically meaningful.

Consistent with recent comparative studies,^5,9 no significant between-group differences were observed in short-term clinical scores or functional performance measures despite the superior accuracy achieved with the Mako system. This may reflect a threshold effect, whereby the outcomes are unlikely to differ once both techniques achieve alignment within the safe zone, and the multifactorial nature of patient-reported outcome measures (PROMs). Moreover, the benefits of greater accuracy are more likely to affect long-term outcomes such as dislocation, wear, or revision rather than early recovery. Robotic assistance has also been reported to enable faster early recovery than CT-based navigation,⁶ although heterogeneity in surgical techniques, comparator systems, and study protocols may explain this finding.

Both operating time and intraoperative blood loss were significantly greater in the Mako group in the present study. These findings suggest potential advantages of AR in surgical invasiveness. However, another study reported minimal differences between the two methods,⁹ and a recent meta-analysis found no significant differences in blood loss or operating time between robot-assisted and conventional techniques,⁴⁰ suggesting that these differences may not always be clinically relevant. Moreover, AR navigation requires additional preoperative steps for pin placement and registration, and the learning curve—particularly for operating time—has been noted in previous reports.^41,42 Thus, the true difference in operating time between Mako and AR may be smaller than that observed during early adoption.

In this study, the AR navigation system demonstrated high cup placement accuracy, with a similarly high rate of placements within Lewinnek’s safe zone as the Mako system. This finding is consistent with previous reports showing comparable or superior accuracy to conventional methods, with similar operating time and blood loss.⁴³ The AR-hip system offers clear practical and economic advantages, as it requires no preoperative CT scans, disposable components, or dedicated console, and can be used at an estimated cost of approximately USD $500 per procedure without maintenance or capital expenses.⁴⁴ In contrast, the Mako system requires an initial investment of $1 million and ongoing operational costs.⁴⁵ Although robotic-assisted THA has been reported to be cost-effective relative to conventional techniques in high-volume centres,^46,47 the AR system can achieve comparable accuracy at substantially lower costs, offering a practical and flexible option for uncomplicated primary THA depending on institutional resources and case complexity. However, formal cost-effectiveness evaluations are currently limited. Future prospective studies incorporating clinical outcomes and comprehensive cost analyses are warranted to clarify the optimal indications and economic value of each system.

This study has some limitations. First, because the analysis was restricted to cases with postoperative CT scans, selection bias may have occurred. The relatively small sample size limited covariate adjustment in the overlap weighting model because including all variables increased the SMDs. Second, the postoperative CT timing varied. Although sensitivity analyses supported the robustness of the results, measurement bias could not be fully excluded. Third, because the systems were introduced at different times, secular trends—such as changes in surgeon experience, device maturity, and workflow optimization—may have favoured the Mako group. Residual confounding factors related to the early phases of device implementation may persist. During the brief overlapping period when both systems were available, the choice of navigation system was based on the surgeon’s preference. However, most AR navigation cases (35 of 45) were performed before the introduction of the Mako system, minimizing potential selection bias related to device availability. Additionally, different implant systems were used between the two groups. Although the primary outcomes were cup alignment and positional accuracy, which are determined mainly by the surgical workflow and navigation guidance rather than implant design, this difference should still be acknowledged as a potential source of bias. Furthermore, as this was a single-surgeon, single-centre study, generalizability is limited. Device-specific characteristics, such as constrained intraoperative adjustments of the robotic arm system compared with manual corrections possible with AR navigation, may also reflect institutional factors. Finally, the follow-up period was limited to the time of discharge. However, this study was designed to evaluate short-term recovery and early functional improvement; further investigations with longer follow-ups are warranted to assess long-term outcomes.

This study also has several strengths: it is the first to directly compare robotic arm assistance and AR navigation in THA with adjustments for confounders. Additionally, overlap weighting of propensity scores was applied to minimize statistical bias. Furthermore, cup placement accuracy was assessed three-dimensionally using postoperative CT, evaluating both angular orientation and COR positioning, and combined with short-term clinical and functional outcomes. These strengths support the reliability of our findings. Robotic arm assistance achieves accurate and reproducible cup placement and may offer advantages in complex anatomy, although its clinical significance requires further validation.

Future research should include multicentre randomized trials comparing accuracy and long-term outcomes, cost-effectiveness analyses to inform adoption, and studies on the impact of learning curves and surgeon-related factors, and evaluate implant survival, effect modification by case complexity, and whether device selection should be tailored to case complexity, anatomy, surgeon experience, and institutional resources.

In conclusion, Mako robotic arm-assisted THA provides superior acetabular component placement accuracy than AR navigation while maintaining comparable short-term clinical outcomes. Both systems demonstrate clinical acceptability, with technology selection potentially guided by case complexity, institutional capabilities, and long-term outcome priorities requiring further investigation.

Take home message

- Robotic arm assistance achieved more accurate cup placement and centre of rotation reconstruction than augmented reality navigation.

- Short-term clinical outcomes were comparable between systems, suggesting that both are clinically acceptable in routine practice.

- Robotic arm assistance may be particularly advantageous in technically demanding cases, such as obesity or developmental dysplasia of the hip.

Data Availability

The data that support the findings for this study are available to other researchers from the corresponding author upon reasonable request.