Success in implant survival after total knee arthroplasty hinges on many factors, and the postoperative alignment of the knee is often considered important to long-term survival. Previous studies have shown the negative effect of excessive postoperative varus or valgus alignment on implants after total knee arthroplasty. Berend et al. reported that a varus tibial component alignment of >3.0° and a varus overall limb alignment of ≤0° were associated with medial bone collapse and failure of the tibial component1. Other studies showed that the likelihood of total knee arthroplasty failure significantly increased in knees with <2.5° (p = 0.0081) and >7.4° (p = 0.0053) of anatomic valgus, and suggested aligning knees within this range2,3. Neutral or slight valgus alignment is the most favorable for long-term implant survival, and properly aligned prostheses are associated with better stability and higher clinical scores4-8. Nevertheless, failure still occurs despite favorable postoperative alignment.
Knees with increased preoperative deformity can often be corrected to neutral during total knee arthroplasty, but it is unclear whether these knees are more likely to fail than knees with normal preoperative alignments. One study9 showed that knees aligned >10° of either varus or valgus preoperatively were more likely to fail than knees neutrally aligned. However, Sorrells et al.9 noted that abnormally aligned preoperative knees were also more likely to have an abnormal postoperative alignment, possibly contributing to failure independent of preoperative alignment.
This study builds on the data from previous studies performed at our center that emphasized the influence of postoperative knee alignment on prosthesis survival after total knee arthroplasty. The primary purpose of the current study was to analyze how preoperative alignment affected the results of prosthesis survival after total knee arthroplasty, to establish specific ranges for what constitutes disadvantageous preoperative alignment, and to evaluate the relationship between preoperative and postoperative knee alignment.
From September 1983 to November 2006, 5342 primary total knee arthroplasties (in 3699 patients) with complete anatomic alignment data and more than two years of follow-up were performed at our center; Anatomic Graduated Component (AGC) implants (Biomet, Warsaw, Indiana) were used in these arthroplasties. Fifty-four (1.0%) of these knees required a revision for reasons not related to infection or isolated extensor mechanism complications. Preoperative and postoperative alignment data were available for all revised knees. Of the remaining knees that did not fail, 728 did not have a preoperative alignment recorded in the database, and 727 did not have a postoperative alignment in the database; these knees were not included in the study group of 5342 knees. The average duration of follow-up (and standard deviation) for living patients who were not lost to follow-up was 7.1 ± 3.7 years (range, two to 22.5 years). Of the 3699 patients (2242 women [60.6%] and 1457 men [39.4%]), the average age (and standard deviation) at the time of the index operation was 70.2 ± 8.5 years (range, twenty-one to ninety-two years); 1064 patients (28.8%) died during the course of the study. The average patient body mass index (BMI) (and standard deviation) recorded just prior to surgery was 30.0 ± 5.5 kg/m2 (range, 16.5 to 64.3 kg/m2). Osteoarthritis was the preoperative diagnosis for 5089 knees (95.3%), rheumatoid arthritis for 178 knees (3.3%), osteonecrosis for sixty knees (1.1%), and other diagnoses for fifteen knees (0.3%); the other diagnoses included hemophilia (four knees), chondrocalcinosis (four knees), ochronosis (four knees), Paget disease (one knee), von Willebrand disease (one knee), and systemic lupus erythemadosus (one knee).
Of the 9624 consecutive knees that underwent total knee arthroplasty at our institution during the time period of the study, 2204 did not fail but were excluded from the study because they had less than two years of follow-up, 482 knees (5.0%) were completely lost to follow-up, and 141 knees were excluded from the study because they had undergone a revision rather than a primary procedure.
All knees received a cemented, posterior cruciate-retaining, AGC total knee replacement (Biomet). The tibial components were made with Himont 1900 polyethylene (Biomet) that was compression molded onto a cobalt-chromium component, producing a nonmodular or monoblock design, and were implanted without augments on the medial or lateral tibial plateau. All knee replacements were performed by one of six surgeons (four of whom were authors in this study: M.A.R., R.A.M., M.E.B., and J.B.M.) using similar instruments and techniques. Ligament releases on the medial side (deep medial collateral, superficial medial collateral, and posterior medial capsule) and the lateral side (iliotibial band, lateral capsule, and lateral collateral ligament) were done sequentially until the surgeon believed that the knee was aligned correctly and was stable, and no medial or lateral tibial augments were used. Each knee was independently measured once again by the operating surgeon before and after surgery for overall coronal alignment. Postoperative knee alignment values for this study were determined by averaging the alignments obtained at each follow-up evaluation. All alignments were measured on short-film radiographs by either the operating surgeon or one of the six orthopaedic surgeons at our center (four of whom were authors in this study: M.A.R., R.A.M., M.E.B., and J.B.M.); the surgeon used a goniometer and measured the femorotibial anatomic alignment axis, expressed in negative values for anatomic varus alignments and in positive values for anatomic valgus alignments. In this article, negative values were converted to positive values in varus and identified by the word “varus” in each instance.
Statistical examinations were performed to determine associations between preoperative deformity and survival after including the additional significant covariates of BMI, postoperative tibial alignment, postoperative femoral alignment, and age. Time to failure data was determined with use of Kaplan-Meier survival analysis. All p values and hazard ratios were given by means of Cox regression with use of SAS 9.3 (SAS Institute, Cary, North Carolina), with significance set at p < 0.05. Failure was defined as aseptic tibial or femoral loosening for most statistical analyses, except those showing the specific failure mechanisms of tibial collapse and ligamentous instability.
The postoperative neutral range of alignment of 2.5° through 7.4° was used because this range, both independently and including the effects of preoperative alignment, showed the lowest rates of failure and revision in a prior study2 that began with the same patient cohort as the present study. The average preoperative anatomic alignment (and standard deviation) was 0.1° ± 7.7° of varus (range, 25° of varus to 35° of valgus), and the average postoperative alignment (and standard deviation) was 4.7° ± 2.5° of valgus (range, 12° of varus to 20° of valgus). The prior study2 showed that knees with neutral alignment postoperatively (2.5° through 7.4°), regardless of preoperative alignment, had a lower failure rate (0.74%) than knees with misalignment postoperatively (1.7%) (p < 0.0001). The preoperative neutral range of alignment was determined by testing every possible interval, in whole-number degrees, occurring in the present dataset (n = 5342) across the entire preoperative alignment range. The algorithm was written in SAS 9.3 with use of a Cox regression that included indicator variables (0 or 1) for patient age of ≥71 years, BMI of ≥41 kg/m2, and postoperative alignment of 2.5° of valgus through 7.4° of valgus. The most significant segment of the range (the range with the lowest p value), showing the largest numerical differences in hazard ratios, was selected to define the neutral range for preoperative alignment in the present study.
Source of Funding
No external source of funding was used in support of this study.
The range of preoperative knee alignment with the best long-term survival was 8° of varus to 11° of valgus (Table I). Knees falling outside of the range of preoperative 8° of varus and 11° of valgus had 4.3 times the risk of failure than knees within the range (hazard ratio, 4.3; p < 0.0001). This Cox regression included the additional effects of postoperative malalignment (<2.5° or >7.4° of valgus), which had a hazard ratio of failure of 2.6 (p = 0.0004) compared with neutral postoperative alignment (≥2.5° and ≤7.4° of valgus), on patients with regard to BMI (hazard ratio, 3.5 [p = 0.0033], when patients with a BMI of ≥41 kg/m2 were compared with those with a BMI of <41 kg/m2) and age (hazard ratio, 4.4 [p < 0.0001], when patients with an age of <71 years were compared with those with an age of ≥71 years). The aseptic all-cause failure rate for knees within 8° of varus to 11° of valgus preoperatively was 0.71% (thirty-one of 4342 knees). The failure rate was 2.2% (hazard ratio, 3.4; p = 0.0005) for knees with preoperative anatomical alignment of >8° of varus and 2.4% (hazard ratio, 2.7; p = 0.00081) for knees with preoperative anatomical alignment of >11° of valgus (Table I).
Failure was considered to be aseptic loosening of the femoral and/or tibial component for any reason; according to this criterion, fifty-four (1.0%) of 5342 knees failed. Of the fifty-four implant failures, twenty-six (48.1%) were caused by tibial collapse, sixteen (29.6%) by ligamentous instability, seven (13.0%) by femoral component failure, and five (9.3%) by aseptic radiographs with radiolucency in all zones about the femoral or tibial prosthesis.
Three thousand one hundred and fifty-one knees were neutrally aligned both preoperatively and postoperatively and had a failure rate of 0.51% (Table II). Of the 635 corrected knees, the failure rate was 1.9% (twelve knees) for knees corrected from preoperative varus (six of 326 [1.8%]) or valgus (six of 309 [1.9%]) to neutral alignment postoperatively, compared with the failure rate of 3.0% (eleven of 365) for knees that were undercorrected, uncorrected, or overcorrected (hazard ratio, 2.9; p = 0.0130) (Table II).
Knees that were in valgus alignment preoperatively (n = 461) were more susceptible to failure if they were in either excessive varus alignment (<2.5° of valgus) (n = 31) or excessive valgus alignment (>7.4°) (n = 121) postoperatively as compared with neutral alignment (n = 309) postoperatively (p = 0.0336; hazard ratio, 3.8; failure rates, 3.3% [five of 152] compared with 1.9% [six of 309]) (Table II). Knees that were in neutral alignment (≤8° of varus and ≤11° of valgus) preoperatively (n = 4342) were more susceptible to failure if they were in either excessive varus alignment (<2.5° of valgus) (n = 795) or excessive valgus alignment (>7.4°) (n = 396) postoperatively as compared with neutral alignment (≥2.5° and ≤7.4°) postoperatively (p = 0.0116; hazard ratio, 2.5; failure rates, 1.3% [fifteen of 1191] compared with 0.5% [sixteen of 3151]) (Table II). With the numbers available, knees that were in varus alignment preoperatively (n = 539) were numerically more susceptible to failure if they were in varus alignment postoperatively (n = 128) as compared with those in neutral alignment postoperatively (n = 326) (p = 0.1096; insignificant hazard ratio, 2.6 [95% confidence interval (CI), 0.8 to 8.8]; failure rates, 3.9% [five of 128] compared with 1.8% [six of 326]), or if they were in valgus alignment postoperatively (n = 85) compared with those in neutral alignment postoperatively (n = 326) (p = 0.6612; insignificant hazard ratio, 1.6 [95% CI, 0.2 to 13.4]; failure rates, 1.2% (one of eighty-five) compared with 1.8% [six of 326]) (Table II).
Tibial collapse and instability were significantly associated with preoperative alignment in different ways for each of these failure modes (Table III). Of the twelve failed knees in varus preoperatively, eight (66.7%) failed because of medial tibial collapse. Knees that were in varus both preoperatively and postoperatively were more likely to fail because of medial tibial collapse than those that were in varus preoperatively but neutral postoperatively (p = 0.0220, hazard ratio, 5.4; failure rates, 3.9% [five of 128] compared with 0.9% [three of 326]. Of the eleven failed knees in valgus preoperatively, eight (72.7%) failed because of lateral ligamentous instability. Knees that were in valgus both preoperatively and postoperatively were more likely to fail because of lateral instability than those that were in valgus preoperatively but neutral postoperatively (p = 0.0073, hazard ratio, 8.3; failure rates, 3.3% [four of 121] compared with 1.0% [three of 309]).
The influence of postoperative knee alignment on implant survival after total knee arthroplasty is a topic of debate. It has been shown that knees aligned in excessive varus or valgus after surgery have a greater rate of failure than neutrally aligned knees1-4,8. Postoperative neutral alignment reduces the bone stresses and strains typically seen with extreme varus or valgus alignments by producing equal force distribution across the tibia7,10. Although the effects of postoperative alignment were examined, the central focus of the present study was how preoperative alignment related to implant survival after total knee arthroplasty; specifically, we quantified the relationship between preoperative alignment and postoperative alignment.
Knees that were aligned outside of 8° of varus or 11° of valgus preoperatively failed at a significantly higher rate than knees aligned within this range, even if returned to neutral (2.5° through 7.4°) postoperatively. The data strongly indicate that preoperative alignment outside of the range of 8° of varus (p = 0.0005) to 11° of valgus (p = 0.0081) increases the probability of total knee arthroplasty failure.
Although preoperative deformity as an influence on failure is uncontrollable, postoperative alignment as an influence can be controlled. The majority of knees in this study, including those with severe preoperative misalignment, were placed in neutral alignment during total knee arthroplasty. Specifically, 70.9% of knees were aligned from 2.5° through 7.4° of valgus postoperatively, which corresponds to the recommended range of alignment in previous reports2,3. Knees that were aligned neutrally both before and after surgery exhibited the best survival rate (0.51% failure rate), and the majority of knees (59.0%) fell into this category. It is important that preoperative neutral knees are aligned in neutral position during surgery, as this group was 2.6 times (p = 0.0006) less likely to fail than those that were neutrally aligned preoperatively but were in varus or valgus postoperatively.
Although knee alignment correction to neutral does improve the prospect of survival for all preoperative alignments, it does not guarantee success. The fact that knees with preoperative anatomical alignment that is >8° of varus or >11° of valgus failed at a higher rate than preoperatively neutral knees even when corrected to neutral postoperative alignment suggests that a large preoperative deformity is a risk factor for total knee arthroplasty failure independent of postoperative alignment. Previous studies have established that increased tibial surface strains are associated with varus or valgus alignments, but these studies all focused on alignments obtained postoperatively1,6-8. Knees with preoperative alignment deformities may experience higher tibial strain similar to that of knees with postoperative malalignment, or they may also be a result of ligament instability.
Although patients outside of the neutral preoperative knee alignment range (8° of varus through 11° of valgus) were at higher risk for total knee arthroplasty failure, the probability for implant survival was significantly improved with a neutral postoperative alignment. Our analysis showed that if extreme valgus alignment was not adjusted to neutral during surgery, the component was 3.8 times more likely to fail (p = 0.0336). Likewise, if excessive varus alignment was uncorrected during total knee arthroplasty, the implant was 2.6 times more likely to fail than if it had been corrected to neutral, although this result was not yet significant (p = 0.1096).
A limitation of this study was that it was a retrospective review, so we did not specifically match patient demographic characteristics such as age, sex, BMI, and diagnosis between knees that failed and those that did not or between classes of preoperative deformity. Instead, we ran the Cox regression to best account for these additional covariate effects. Although the size of the examined cohort was large, smaller numbers in the subdivided preoperative and postoperative alignment groups rendered some statistical comparisons insignificant. These comparisons would be strengthened with the addition of extra years of data, but our present argument appears to be well supported by the data, considering the many significant differences found among the various failure rates. Concern could also be raised with the absence in the database of preoperative and postoperative alignments for some knees in the cohort of knees that did not fail. To demonstrate that these missing values were not detrimental to our conclusions, a sample of those knees with missing data in the database (n = 79) was taken from archival charts. The mean alignment distribution (and standard deviation) of this sample (0.5° ± 7.8°) was similar to that of the successful knees in the cohort (0.1° ± 7.7° of varus) with regard to the results of the two-sample t test (p = 0.4963) and the Kuiper test for differing distributions (p = 0.5172) and was significantly different with regard to the chi-square test of the comparison between the sample group and failed implants (p = 0.0124) (Fig. 1). Because the distributions of the cohort of knees that did not fail and the archival random sample of knees that also did not fail were similar, and the comparison between the distribution of the archival random sample and the knees that failed was significantly different (p = 0.0124), we concluded that lost or missing data on preoperative alignment for knees that did not fail after total knee arthroplasty did not unduly influence the results of the present study. We did not evaluate interrater discrepancies among the assessors of alignment based on low interrater variability during past studies at our institution1-4.
Knees aligned outside of the range of 8° of varus through 11° of valgus before surgery were more likely to require a revision. Although aligning the knee within 2.5° through 7.4° postoperatively significantly improved the likelihood of implant survival, preoperative deformity by itself increased the risk of implant failure after total knee arthroplasty with regard to preoperative varus (2.2%; p = 0.0005) and preoperative valgus (2.4%; p = 0.0081). Poor preoperative alignment and poor postoperative alignment were associated with a higher risk of failure after total knee arthroplasty. The surgeon should pay close attention to alignment in surgery, especially when the knee is in severe preoperative varus or valgus. Uncontrollable preoperative deformities will always exist in some patients, but postoperative alignment can be controlled by careful and accurate surgical procedure.
Note: We thank E. Michael Keating, MD, and Phillip M. Faris for contributing cases to this study.
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. None of the authors, or their institution(s), have had any financial relationship, in the thirty-six months prior to submission of this work, with any entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. Also, no author has had any other relationships, or has engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.