Total Knee Arthroplasty Component Templating

Total Knee Arthroplasty Component Templating

The Journal of Arthroplasty Vol. 27 No. 9 2012 Total Knee Arthroplasty Component Templating A Predictive Model Adam G. Miller, MD,* and James J. Purt...

79KB Sizes 2 Downloads 150 Views

The Journal of Arthroplasty Vol. 27 No. 9 2012

Total Knee Arthroplasty Component Templating A Predictive Model Adam G. Miller, MD,* and James J. Purtill, MDy

Abstract: Preoperative planning is essential to total knee arthroplasty (TKA); however, TKA templating is historically inaccurate. To improve on templating accuracy and streamline preoperative planning, we set out to predict component sizes based on patient characteristics without radiographs. A total of 123 consecutive patients undergoing unilateral TKA were identified and included in the model study. Input variables consisted of age, gender (as a binary number), height, weight, and body mass index. A linear regression model was created. The models predicted component size exactly in 74% of femurs and 85% of tibias. All model predictions were within a ±1 size of the actual components implanted. Our models were more accurate than any previous model for TKA reported. Keywords: total knee arthroplasty, templating, preoperative planning. © 2012 Elsevier Inc. All rights reserved.

Preoperative planning is essential to total knee arthroplasty (TKA). With appropriate planning, implants have improved survival [1], and the duration of surgery is shorter [2]. Crucial to thorough planning are appropriate weight-bearing roentograms. These radiographs are assessed for alignment, previous pathology or trauma, and component sizing. Preoperative templating attempts to accurately determine the size of femoral and tibial components. A surgeon will then know what sizes are required to be available for the operating room and the most likely starting point during intraoperative sizing. Previous studies have investigated preoperative templating using acetate and digital techniques to determine component sizes required [3-9]. Templating of femoral and tibial components has been reported to be accurate approximately 50% to 60% of the time. Component sizing accuracy within a ±1 size of the actual components implanted averages 90% to 95%. The difference between digital and analog templating accuracy is equivocal. These studies suggest that current methods

of preoperative templating should only be used as a guide for initial trial components used. Furthermore, it can only reliably predict the size of implants needed within 2 sizes accurately. Most imaging has moved to digital storage and viewing. Although no benefit to analog or digital templating has been clearly shown, templating will likely continue to move digital as well. With this transition, further confusion without consensus revolves around proper magnification of digital images for accurate templating. Multiple methods exist, with no clear benefit of one vs another. In an attempt to improve on templating accuracy, remove templating confusion, and streamline preoperative planning, we set out to predict component sizes based on patient characteristics. This approach may provide an objective measure of component selection that can be made independently of radiographs. A predictive model was created and tested against the accuracy of previous methods.

From the *Department of Orthopedics, Thomas Jefferson University Hospital, Curtis Bldg 801, 1015 Walnut St, Philadelphia, Pennsylvania; and yRothman Institute, Philadelphia, Pennsylvania. Submitted September 23, 2011; accepted March 28, 2012. Funding or grants received: None. Institutional review board: Not applicable. The Conflict of Interest statement associated with this article can be found at doi:10.1016/j.arth.2012.03.055. Reprint requests: Adam Miller, MD, 1015 Walnut St, Curtis Bldg Rm 801, Philadelphia, PA 19107. © 2012 Elsevier Inc. All rights reserved. 0883-5403/2709-0021$36.00/0 doi:10.1016/j.arth.2012.03.055

Patients who underwent primary cemented TKA were identified retrospectively by the senior author in 2009 and 2010. A total of 123 consecutive patients undergoing unilateral TKA were identified and included in the model study. This study was approved by the institutional review board's committee approval board. Data collected included age, gender, height, and weight. Operative logs were reviewed for femoral and tibial component type and sizing implanted at the final cementation. All implants were cemented to distal

Materials and Methods

1707

1708 The Journal of Arthroplasty Vol. 27 No. 9 October 2012 femur and proximal tibia with applied tourniquet. Depuy PFC Sigma (Depuy Orthopaedics, Warsaw, Indiana) components were used. The manufacturer's intramedullary distal femoral guide and anterior-posterior sizing guide were used for preparing the femur and sizing the femoral component. Extramedullary tibial cutting guide and tibial plateau templates were used in tibial component sizing. Predictive models were created. Input variables consisted of age, gender (as a binary number), height, weight, and body mass index (BMI). Dependent output variables (femoral and tibial component sizes) were assessed in 2 separate models. Patient data were initially assessed for normal distribution using the KolmogorovSmirnov test. Pearson product-moment coefficients were then used to determine the relationship between patient variables and dependent outcome variables. A linear regression model for both femur and tibia was created based on acquired variables from these 123 patients, and the accuracy of this model was assessed for the patients by determining how many predicted sizes fell within 2 sizes or 1 size, or were exact. The first variable added to the model was the variable with the largest Pearson coefficient. Thereafter, variables were sequentially added to the model in order of their Pearson coefficient and only kept if the adjusted R 2 increased significantly as determined by the F statistic. After multivariate regression, variables were removed from the model if they were nonsignificant (F N 0.05). Predicted sizes were considered exact matches if the difference between the actual size and the predicted size was less than 0.5 size unit. Predicted sizes that differed 1.5 size units were considered within 1 size. Similar measurements were made for within 2 size units. This model was then validated internally by applying the predictive model to 60 patients who underwent primary cemented TKA from the same surgeon who were not in the group of 123. Accuracy was then calculated in a similar manner.

Results Of the 123 patients, there were 80 women and 43 men having an average age of 64.2 years (range, 33-86 years). Average height, weight, and BMI were 65.8 in, 202.7 lb, and 32.73 lb/in 2, respectively (Table 1). The Table 1. Patient Data Variable

Mean

SD

Tibia sizes Femur sizes Age (y) Height (in) Weight (lb) BMI (lb703/in²)

2.92 3.04 64.17 65.86 202.73 32.73

0.79 0.91 9.22 4.14 42.96 5.64

Patient data based on 123 patients with unilateral total arthroplasty.

range of Depuy PFC Sigma femur sizes implanted was 1.5 to 5. The same was true for the tibial components. Linear regression was performed, and significant variables were identified. For the femoral sizing prediction model, height, gender, and age were predictive of component sizing. This yielded the following equation for the final component size: femur size = − 6:348 + 0:137ðheightÞ − 0:533ðgenderÞ + 0:011ðageÞ where female is given a value of 1, and male, zero. Height is measured in inches, age is in years, and weight is measured in pounds. The model returned an adjusted R 2 of 72% (Table 2). For the tibial sizing prediction model, height, gender, and age were predictive of component sizing. This yielded the following equation for the final tibial component size: tibia size = − 3:96 + 0:011ðageÞ − 0:639ðgenderÞ + 0:0929ðheightÞ + 0:00233ðweightÞ where, again, female is given a value of 1, and male, zero. The model returned an adjusted R 2 of 79% (Table 2). The models predicted component size exactly in 74% of femurs and 85% of tibias. All model predictions were within a ±1 size of the actual components implanted 100% of the time. For both models, height was the largest predictor of component size, followed by gender and age. Body mass index was only predictive in the tibia model for this study but was not added due to a lack of independence in the model. Weight was predictive in the femur model only. These predictive models were then applied to a separate series of 60 patients for validation of the models. There were 32 women and 29 men having and average age of 65.7 years (range, 50-86 years). Average height, weight, and BMI were 67.2 in, 202.7 lb, and 31.4 lb/in 2, respectively. When the models were applied, femoral component sizing was exactly predicted 73.3% of the time and tibial component sizing 85% of the time, and all model predictions were within 1 size of the actual components implanted.

Discussion Planning for a TKA is crucial. Comorbidities are identified, perioperative issues are addressed, and appropriate surgical planning occurs. Preoperative

Table 2. Prediction Models Statistics Model

R2

Femur Tibia

0.725 0.794

Adjusted R2

SE of the Estimate

0.718 0.787

0.4819 0.3646

Linear regression final model data. Femur predictors: height, gender, and age. Tibia predictors: height, gender, age, and weight.

Total Knee Arthroplasty Component Templating  Miller and Purtill

templating is one aspect of surgical planning. The benefit of templating in TKA surgery appears to be 2-fold. First, the surgeon is able to reliably predict a range of implant sizes needed to be available in the operating room. Second, templating provides the surgeon with a reliable starting point in determining implant size and position without making aberrant cuts during the initial sizing. One then knows which sizes should be available for use. Both benefits may decrease surgical time and, therefore decrease complications. Previous studies have results of analog and digital templating methods with no clear superior method. Furthermore, they show equivocal efficacy in accurately predicting exact sizes of implemented components [7]. We present a novel model for “templating” component sizes that bypasses all concerns of sizing methods. In addition, our models were more accurate than any previous model for TKA reported. The ability to be accurate 100% of the time within 1 size of the actual components may have implications in operating room stocking and instrument preparation. Although this model allows improved size prediction, we still advocate all sizes to be available on site. In addition, this improved implant size prediction model may facilitate femur and tibia preparation that is more consistent with the appropriate final component sizing. This would appear to save time by decreasing the need to revise an initial sizing plan. There are limitations to this study. Although we have demonstrated a predictive model, this model will not be transferrable to other component types and designs. This model is validated specifically for Depuy Sigma components and this particular surgeon. However, the process and variables analyzed create a model that could possibly be reengineered for individual product lines and other surgeons, given a thorough external validation study.

1709

In conclusion, mathematical models based on linear regression of easily obtained nonidentifying information about a patient provide a more accurate template of TKA sizes. Implementation of such a model may streamline preoperative planning and reduce intraoperative time. Further investigation toward more generalizable models would be needed for implementation and broader validation.

References 1. Schwartz JT, Mayer JG, Engh CA. Femoral fracture during non-cemented total hip arthroplasty. J Bone Joint Surg 1989;71-A:1135. 2. Blackley HR, Howell GE, Rorabeck CH. Planning and management of the difficult primary hip replacement: preoperative planning and technical considerations. Instr Course Lect 2000;49:3. 3. The B, Diercks RL, van Ooijen PM, et al. Comparison of analog and digital preoperative planning in total hip and knee arthroplasties. A prospective study of 173 hips and 65 total knees. Acta Orthop 2005;76:78 [PubMed PMID: 15788312]. 4. Specht LM, Levitz S, Iorio R, et al. A comparison of acetate and digital templating for total knee arthroplasty. Clin Orthop Relat Res 2007;464:179 [PubMed PMID: 18062050]. 5. Aslam N, Lo S, Nagarajah K, et al. Reliability of preoperative templating in total knee arthroplasty. Acta Orthop Belg 2004;70:560. 6. Del Gaizo D, Soileau ES, Lachiewicz PF. Value of preoperative templating for primary total knee arthroplasty. J Knee Surg 2009;22:284. 7. Howcroft DW, Fehily MJ, Peck C, et al. The role of preoperative templating in total knee arthroplasty: comparison of three prostheses. Knee 2006;13:427. 8. Unnanuntana A, Arunakul M, Unnanuntana A. The accuracy of preoperative templating in total knee arthroplasty. J Med Assoc Thai 2007;90:2338. 9. Trickett RW, Hodgson P, Forster MC, et al. The reliability and accuracy of digital templating in total knee replacement. J Bone Joint Surg Br 2009;91:903.