Introduction

Total knee arthroplasty is widely performed for disabling arthritic knees. Main surgical goals include correcting limb alignment with proper balance and deformity correction.¹ Over the decades, multiple alignment techniques have been developed for knee arthroplasty, and can be broadly classified into three categories:²

• Systematic alignment: Restores neutral alignment i.e. 180o of hip-knee-ankle axis such as mechanical alignment (MA).
• Patient-specific alignment: Restores patient specific pre-arthritic limb alignment and joint line obliquity such as kinematic alignment (KA)
• Hybrid alignment: A blend of both principles such as adjusted mechanical alignment (aMA).

Mechanical alignment (MA) is considered the ‘classic alignment’ and is widely used, in which neutral alignment is believed to be essential for implant survivorship and satisfying patient outcomes.^3–6 It creates a balanced force distribution for medial and lateral compartments to reduce wearing and loosening.^6,7 Nevertheless, despite numerous advancements in implant design and cutting guidance such as robot assisted arthroplasty, Bourne et al reported up to 19% of patients not satisfied with their prosthetic knees.⁸

In response to such data, a novel approach of adjusted mechanical alignment (aMA) that develops on top of classical mechanical alignment has been proposed. aMA uses systematic approach from mechanical alignment for tibial cut and a patient-specific adjustment for distal femoral cut allowing for up to 3^o deviation from neutral axis to achieve equal ligament tension with less ligament release required than classical MA.⁹ It follows newer school of thought that non-neutral alignment does not affect clinical outcome of the implant¹⁰ and the resulting implant positioning may better match patient-specific knee anatomy, resulting in greater patient satisfaction.

Hommel et al⁹ compared aMA and MA techniques showing improved clinical outcomes for knee society score (KSS)¹¹ [93.2 +/- 3.2 (SD) vs 91.1 +/- 3.8 (SD)], Western Ontario and McMaster Universities index (WOMAC score)¹² [19.7 +/- 3.1 (SD) vs 22.0 +/- 3.8 (SD)], and forgotten joint score (FJS)¹³ [58.6 +/- 7.3 (SD) vs 52.8 +/- 8.7 (SD)]. However, it is not clear whether this is consistent in the Asian group, and Moser et al show that Asians have statistically greater femoral mechanical angle and lower tibial mechanical angle, showing Asians have different alignment characteristics than Caucasians.^14,15 Thus, the purpose of this study is to investigate and compare the outcomes of neutral versus adjusted mechanical alignment in total knee arthroplasty in an Asian population. We hypothesize that adjusted mechanical alignment will provide a better clinical outcome for Asian population.

This research will be conducted in compliance with regulatory requirements by International Council for Harmonisation Guideline for Good Clinical Practice (ICH-GCP) and Declaration of Helsinki.

Methodology

Patients

Patients with total knee arthroplasty (TKA) performed within 2020 at a single centre (Total Joint Replacement Centre, Yan Chai Hospital, Hong Kong) and mechanical alignment utilised were reviewed (n=351). Inclusion criteria was patients with primary diagnosis of osteoarthritis who underwent total knee arthroplasty. Exclusion criteria included prior knee surgery (n=0), valgus deformity (n=14), incomplete preoperative and/or postoperative data (n=116), and non-osteoarthritis primary diagnosis such as post-traumatic arthritis, rheumatoid arthritis (n=2). Patients were separated into neutral mechanical alignment group (MA) with neutral axis adopted intra-operatively (n=125) and adjusted mechanical alignment group (aMA) with up to 3^o varus from femoral cut adopted (n=62).

In the database, each TKA was considered a separate entry. Entries from patients with bilateral TKA (n=70) done in the same session had the same baseline characteristics, which may produce overlapping data should each side of the TKA belonged to a different group (i.e. one in MA group and other side in aMA group). Moreover, whilst such patients provided clinical outcome scores of both sides of TKA in separate questionnaires, potential reporting bias still existed. To fulfil the statistical assumption where each observation is independent and to reduce potential bias, nested sampling was performed in R using a fixed random seed such that each patient contributed only to one TKA entry should they had both legs operated under the same alignment (n=56) or whole entry omitted if both legs were operated under different alignment concepts (n=14) (Figure 1).

Figure 1

Study flowchart

Data

Patient demographics including age, gender were recorded. Preoperatively, standing whole lower limb radiographs were obtained with mechanical femoral-tibial angle (MFTA), anatomical femoral-tibial angle (AFTA) measured by same observer. Range of motion, Knee Society Score (KSS) with knee and function sub-scores, and WOMAC score were documented. Postoperatively all patients were followed up for up to 1 year, and range of motion, KSS knee score, KSS function score, WOMAC score, and Forgotten Joint Score (FJS) were measured. Range of motion was defined as a single number representing the difference in angle between maximal knee extension and flexion.

Additional information regarding postoperative complications including iatrogenic fracture, postoperative wound infection, thromboembolism, revision surgery, and mortality was also recorded.

Propensity score matching

To reduce confounding bias between MA and aMA groups, propensity score matching was performed using R package MatchIt16 in the optimal target ratio of 2:1 without replacement as suggested by Austin et al, with a caliper width equal to 0.2 standard deviations.¹⁷ The nearest neighbor matching algorithm was used, and logistic regression was used to estimate the propensity score. For variable selection, univariate logistic regression was performed using glm in R on baseline variables against treatment assignment to assess their associations.

Variables selected for the final propensity score matching included age, sex, MFTA, AFTA, preoperative active range of motion (AROM), preoperative passive range of motion (PROM), preoperative KSS, and preoperative WOMAC score.

The balance of baseline covariates of matched samples was assessed using absolute standard mean difference. Previous literature suggested acceptable cut-offs of below 0.1 or 0.25.¹⁸

Data analyses

The normality of variables was analysed using Shapiro-Wilk test. Pre- and postoperative ROM, WOMAC scores, and postoperative KSS and FJS were significantly different from normal distribution. Age, MFTA, and AFTA followed a normal distribution.

Summary statistics for baseline characteristics were presented in the format of mean [standard deviation] for normally distributed variables, median [interquartile range] for non-normally distributed variables, and number [percentage] for categorical variables.

Student’s t-test was used to compare mean difference at baseline for normally distributed variables. Wilcoxon rank-sum test was used to compare differences in median for non-normally distributed variables. Chi-square test was used to compare categorical variables.

P-value <0.05 was considered statistically significant unless otherwise specified. All statistical analyses were performed in R.

The primary outcomes of interest were postoperative range of motion, clinical scores including KSS, WOMAC score and FJS comparison, while as the secondary outcome was postoperative complication rate. Confidence interval was not presented due to the use of non-parametric statistical tests.

Results

Table 1

Baseline characteristics before and after propensity score matching

58 of 62 patients in the aMA group were matched to 93 of 125 patients in the MA group. The target matching ratio of 2:1 was not attained due to caliper width limits. Statistical tests were applied as described to compare baseline variables between the treatment groups. P-values < 0.1 were italicized, and p-values < 0.05 were bolded. After propensity score matching, none of the baseline variables were significantly different between the two treatment groups (Figure 1)(See Supplementary Materials figure 3 & 4).

Table 2

Logistic regression between baseline variables with treatment assignment and clinical outcomes

Statistical tests were applied as described to evaluate the associations between baseline variables and treatment assignment, postoperative clinical scores, and postoperative range of motion respectively. P-values < 0.1 were italicized, and p-values < 0.05 were bolded. The p-values were concordant with those obtained by different statistical tests for baseline characteristic comparisons. Although no conventional statistical tests were available to demonstrate clear association between age or preoperative WOMAC score with treatment assignment, these variables were included in the matching due to their potential clinical relevance.

Table 3

Comparison of primary clinical outcomes between MA and aMA group

Figure 2

Wilcoxon Rank Sum test of matched patients

There were no statistically significant differences between MA and aMA groups for postoperative KSS, WOMAC score, FJS, AROM, or PROM.

Table 4

Complications

1 case of iatrogenic medial epicondyle and medial collateral ligament (MCL) avulsion occurred in MA group, which was repaired with cannulated screw and sutures; while 1 case of patellar tendon bony avulsion occurred in the aMA group, which was repaired with suture anchors. Both cases underwent normal rehabilitation postoperatively. There were no other associated surgical complications reported.

Discussion and limitations

We showed that when comparing MA and aMA groups, there is no statistically significant difference in outcomes up to 1 year post-operatively.

To date, MA is still considered ‘gold standard’ by most given its reproducibility with a systematic approach, with multiple previous clinical studies supporting this ideology and relative ease of execution.¹⁹ However, the ‘one size fits all’ neutral mechanical axis to be achieved for good clinical outcomes has been debated in recent years. MacDessi et al introduced the Coronal Plane Alignment of the Knee (CPAK) classification, where knee phenotypes have been described in terms of two independent variables: arithmetic hip-knee-angle for varus/neutral/valgus alignments and also joint line obliquity for apex distal/neutral/apex proximal subgroups.²⁰ It has been reported that up to 30% of arthritic patients are constitutionally varus and 67% have distal apex joint line obliquity, while only 15% have both neutral mechanical axis and joint line obliquity. MA can only match native anatomy for a small proportion of the population, while being an abnormal alignment instead for most.^1,21,22

Moreover, Blakeney et al demonstrated that standardized bone cuts may introduce flexion and extension gap imbalance in up to 36% (using posterior femoral condyle as reference) or 51% of patients (using trans-epicondylar axis as reference).²³ To address such imbalance, soft tissue release is often needed to achieve standardized bone cuts, which can be technically demanding or even lead to further imbalance. However, for varus knees with less than 15^o deformity, soft tissue contracture is rarer than ligamentous laxity.²⁴ Further soft tissue release on an already lax ligament may be unnecessary.

On the other hand, aMA allows under-correction of varus deformities. More equal ligament tensioning are produced by adjusting bone cut instead of excessive soft tissue release, preventing excessive laxity or the potential of further gap imbalance.⁹ Recent literatures shows that such residual varus alignment does not negatively impact prosthesis survival rate nor postoperative pain and functional outcomes, contrary to classical beliefs that advocate importance of MA.^10,25  While this study is unable to confirm our hypothesis, it corresponds to the findings of these literature and further validate this belief that aMA provide comparable instead of inferior outcomes when compared to MA. It suggests aMA to be a viable alternative to MA.

The arbitrary 3^o residual deformity allowance is based on previous literature quoting within 3^o as a well-aligned axis.⁶ Moreover, Hsu et al reported that the mean mechanical hip-knee-ankle angle in Asians is -1.2^o (SD 3.2, negative = varus), thus a 3^o residual deformity may be adequate for Asian population.¹⁴ In fact this degree has recently also been challenged. Vanlommel et al demonstrated that slight residual varus between 3^o-6^o has the best clinical outcome instead ²⁶ suggesting a larger figure for residual deformity may be beneficial in aMA. Yet, optimal degree of allowance for residual varus deformity is subject to further study and outside the scope of this discussion.

In terms of hybrid alignment, one alternative is restricted kinematic alignment (rKA). It is based on kinematic alignment where distal femoral bone cut is suited for kinematic alignment with adjustments made over tibial bone cut.³ Kinematic alignment itself is a relatively new concept with 3D orientation for implants required. While both are hybrid techniques, we prefer aMA over rKA and thus studied aMA versus MA for the following reasons. First, aMA is developed from the classical MA concept which is widely validated and utilized. Second, aMA takes a single plane into consideration while rKA is 3D oriented, which will have a theoretically shorter learning curve and is technically less demanding. Third, Winnock de Grave et al showed that TKA with rKA and aMA have comparable clinical outcomes with a 1 year follow up, thus evidence lacks to support rKA over aMA. ²⁷

This study has several limitations. First, as a retrospective cohort study, it is prone to selection bias when patients are allocated to MA or aMA group. Whilst acknowledging the effect of reducing sample size and thus the power of this study, we decided to perform propensity score matching to reduce said bias, providing more comparable data for analysis. When building model for propensity score matching, baseline variables such as KSS and WOMAC score may not be independent of each other. This may violate the assumptions in logistic regression for calculating propensity scores. However, since preoperative clinical scores are likely major determinants of postoperative clinical scores, we decided to include both clinical scores in our final model. Second, other diagnoses such as rheumatoid arthritis or post-traumatic arthritis are frequently associated with more soft tissue and bony deformities requiring more advanced soft tissue release or bone cuts, in such circumstances the results of this study may not be applicable and are subject to further study. Third, one of the primary outcomes is KSS, in which the knee sub-score takes into account of neutral alignment postoperatively, which intrinsically results in lower score for aMA group, producing systematic bias. Nevertheless, given the potential under-estimation of aMA group results under KSS scoring system, all other clinical score comparisons showed consistent findings thus validating the overall results. Fourth, we did not perform analysis with regard to degree of preoperative varus deformity. Preoperative degree of varus deformity may influence amount of correction performed, if any, thus have potential clinical relevance in this study.²⁶ To our best knowledge there has yet to be similar study comparing MA and aMA technique in the Asian population, thus additional long-term studies including analyses of preoperative varus deformity are necessary to further validate the results of this study or to determine whether aMA has any potential clinical benefits compared with MA.

Conclusion

Adjusted mechanical alignment for total knee arthroplasty has comparable short-term clinical outcomes with mechanical alignment in terms of range of movement and clinical scores. It is applicable to Asian population with no short-term adverse effects.

Conflict of Interest

The authors declare that they have no competing interests.

Funding

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

Supplementary materials

Figure 3

Exploratory data analysis of the distributions of measured variables

The distribution of each variable was plotted as a histogram, with black vertical line indicating mean and red indicating median. The greater the difference between mean and median was, the more skewed the data were. The p-value of Shapiro test for normality is plotted at the top left corner in each panel. Age, MFTA, and AFTA were normally distributed.

Figure 4

Balance of covariates in propensity score matched samples

Most covariates had an absolute standard mean difference below 0.05, except for AFTA and preoperative PROM, which had values of 0.09 and 0.07 respectively. Previous literature suggested acceptable cut-offs of below 0.1 or 0.25.¹⁸ Therefore, the balance of covariates in the propensity score matched samples was satisfactory.

Figure 5

Distribution of propensity scores

Propensity score was calculated by performing logistic regression on treatment assignment against selected covariates. This could be interpreted as the probability of a patient being assigned to the aMA group given the individual’s baseline characteristics. Patients in aMA group were matched to patients in MA group with similar propensity score. Unmatched patients were not included in this analysis.