The Journal of Arthroplasty 30 (2015) 206–209
Contents lists available at ScienceDirect
The Journal of Arthroplasty journal homepage: www.arthroplastyjournal.org
Association Between Body Mass Index Change and Outcome in the First Year After Total Knee Arthroplasty Alasdair Mackie, BSc(Hons), MBBS a, Karthikeyan Muthumayandi, MPTh, MPhil a, Mark Shirley, BSc, (Hons), PhD b, David Deehan, MD, MSc, FRCS (T&O) a, Craig Gerrand, MB, ChB, FRCSEd (Tr and Orth), MD a a b
Freeman Hospital, Newcastle upon Tyne, United Kingdom School of Biology, Ridley Building, University of Newcastle, Newcastle upon Tyne, United Kingdom
a r t i c l e
i n f o
Article history: Received 21 May 2014 Accepted 2 September 2014 Keywords: weight change BMI TKA impact knee outcome
a b s t r a c t There is an association between obesity, osteoarthritis and total knee arthroplasty (TKA), but little is known about how postoperative weight change influences outcomes. Primary TKA patients were identified from an institutional arthroplasty registry. BMI and patient reported outcome measures (PROMs, specifically WOMAC and SF36) were recorded for 1545 patients preoperatively and up to 3 years postoperatively. Mixed effects modelling showed postoperative BMI change had no impact on postoperative WOMAC scores. However, weight gain over 10% had a negative impact on SF36 pain and functional scores although postoperative weight loss was not associated with improved PROMs. Men showed greater improvement in postoperative SF36 function and pain scores, whilst older patients were slower to improve. Postoperative weight gain has a negative association with SF36 pain and function. © 2014 Elsevier Inc. All rights reserved.
High body mass index (BMI) is associated with higher prevalence of knee osteoarthritis (OA) and reduced reported mobility [1]. Furthermore, it has been linked to poorer outcomes after total knee arthroplasty (TKA) [2,3] including reduced survival because of aseptic loosening [4]. The crucial barriers to weight loss after surgery remain unclear [5]. Increases in BMI in the general population mean there is an increasing awareness of the need to manage weight and interest in how this might be achieved [6]. When patients undergo knee arthroplasty there is a timely opportunity to address comorbidities including obesity. However, there remains little evidence-based guidance for such patients after knee arthroplasty. Given the recognised associations between BMI, knee arthritis and surgical outcomes, we were interested in exploring the impact of BMI change after surgery on patient reported outcomes in an institutional registry, and what advice we might therefore offer patients. Specifically, we were interested in how postoperative BMI change is related to SF36 and WOMAC scores at a year after surgery. Our null hypothesis was that a significant change in postoperative BMI (arbitrarily taken as ±10% body weight or BMI) would not influence patient reported outcomes.
The Conflict of Interest statement associated with this article can be found at http://dx. doi.org/10.1016/j.arth.2014.09.003. Reprint requests: Alasdair Mackie, BSc, Hons, MBBS, 46 Linden Road, Gosforth, Newcastle-Upon-Tyne, Tyne and Wear, NE3 4HB, United Kingdom. http://dx.doi.org/10.1016/j.arth.2014.09.003 0883-5403/© 2014 Elsevier Inc. All rights reserved.
Patients and Methods The Freeman Joint Registry (FJR) was set up in July 2003 as an ongoing institutional audit of patient outcomes following hip or knee arthroplasty. It is registered with and has approval from the institutional research and development board (Project ID number: 3290). Inclusion in the FJR requires informed consent preoperatively for the collection, storage and analysis of data. The study was conducted in accordance with the declaration of Helsinki and the guidelines for good clinical practise. This study was therefore a retrospective comparative cohort study from a single centre institutional arthroplasty registry using anonymised data. Data items used in this study were preoperative patient demographics including; age, gender, comorbidities and self reported height and weight. Patient reported outcomes included the Western Ontario and McMaster University Osteoarthritis Index (WOMAC), and Medical Outcomes Trust Short Form-36 (SF36) [7–10]. Preoperative assessment was undertaken within 6 weeks of surgery and postoperative analysis was performed annually out to 3 years using both WOMAC and SF36 scores. Years 1 and 2 data are collected in out patients; year 3 data are collected using a postal survey. For the purposes of this analysis, we selected patients who underwent primary TKA over a five year period from April 2004 to March 2009 to allow for the collection of follow up data. All patients underwent cemented TKA using either Press Fit Condylar (PFC) (Depuy, Warsaw, Indiana, USA) or Triathlon (Stryker, Marwah, New Jersey, USA) knee implants. No distinction was made between these
A. Mackie et al. / The Journal of Arthroplasty 30 (2015) 206–209 Table 1 Explanatory Variables Used in the Models. Fixed Effects Sex Age BMI preop SF36-FUNC preop SF36-PAIN preop W-FUNC preop W-PAIN preop 10% change BMI preop
Year of assessment Random Effects ID Age category
Definition Categorical Continuous Preoperative body mass index; continuous Preoperative SF36 physical function; continuous Preoperative SF36 bodily pain; continuous Preoperative WOMAC functionality; continuous Preoperative WOMAC pain; continuous Change in BMI from preoperative assessment, divided into three categories: gain (of ≥10% preoperative BMI), loss (of ≥10% preoperative BMI), no change (b10% gain or loss) Ordinal Definition Anonymised identification code; categorical 30–34.9 35–39.9, …, 90–94.9; ordinal
implants are both are modern generation cruciateretaining minimally constrained. All implants were cemented in place. We used linear mixed effects models (LMEs) for this analysis to investigate the role of key variables (Table 1) in explaining the variation in each of the response variables. LMEs were used because they account for repeated measures on the same individuals by including the unique ID of each patient as a random effect. Each unique ID is fitted in the model with a different gradient; with the aim of reducing the residual error of the model by allowing each individual to improve at his or her own rate, rather than assuming that everyone improves at the same rate. All analyses were performed using the 'nlme' package in the R statistical language [11,12], and pseudo-R-squared values generated using the 'MuMIn' package [13]. Marginal R2 calculated by the latter represents the variance explained by fixed effects, whereas conditional R2 is interpreted as the variance explained by both fixed and random effects. Variance attributable to the random effect is therefore the difference between the two. Four response variables were considered in the model: improvement of function score and improvement of pain score from the preoperative assessment, using two different scales: the Western Ontario and McMaster Universities Arthritis Index (WOMAC) and the Short Form (36) Health Survey (SF36). High WOMAC scores reflect poorer outcomes, whereas higher SF36 scores are associated with better outcomes. The following explanatory variables (Table 1) were identified; age, preoperative BMI, sex, SF36 and WOMAC pain and function, and postoperative weight loss or gain (greater or equal to 10%). Percentage weight change was calculated by the difference of postoperative BMI in relation to the preoperative value for each individual patient.
207
There was no significant increase in the rate of in SF36 function score improvement in patients who lost 10% or more of preoperative BMI over those whose weight did not change postoperatively. However, patients who gained 10% or more of their preoperative BMI following total knee arthroplasty had significantly smaller improvements in physical function (t = −2.67, P = 0.008, Fig. 1). Increasing age and increasing preoperative BMI were associated with slower improvements in the SF36 function score (t = − 5.71, P ≤ 0.001 and t = − 4.25, P ≤ 0.001). Male patients had a significantly larger improvement in SF36 function score than female patients (t = 3.52, P ≤ 0.001). Recovery of SF36 function score was positively related to preoperative assessment of pain (t = 7.02, P ≤ 0.001), and negatively associated with preoperative SF36 function score (t = − 13.83, P ≤ 0.001). Overall, the random effect of patient ID accounted for 58.5% of the variance in the model (conditional R2 = 0.704, marginal R2 = 0.119). Improvement of WOMAC function score was significantly related to preoperative BMI (t = − 2.13, P = 0.033) and preoperative WOMAC function score (t = −15.74, P ≤ 0.001). Both these associations were negative, indicating that patients who had a high preoperative BMI and/or a high preoperative WOMAC function score showed a smaller improvement in WOMAC function score than patients with lower preoperative BMI or WOMAC function scores (Fig. 2). The random effect of patient ID accounted for 64.9% of the variance in the model (conditional R 2 = 0.782, marginal R2 = 0.133). Individuals who gained more than 10% of their preoperative BMI following a total knee arthroplasty showed a significant reduction in the rate of improvement of SF36 pain score (t = − 2.58, P = 0.01) compared to those whose weight remained the same or decreased. Weight gain of less than 10% of preoperative BMI was not associated with less improvement in SF36 pain (Fig. 3). Unlike improvements in SF36 function score, age was not a significant predictor of improvement in SF36 pain score. Male patients had a significantly larger improvement in pain score than female patients (t = 3.20, P = 0.001). Preoperative BMI was negatively associated with the rate of improvement (t =
Results There were 1902 patients in the original dataset, however, after accounting for missing data on variables of interest (age, sex, BMI, PROMS for pain and functionality) there were records for 1821 individuals. Of these, 276 without follow up data were excluded, leaving a sample size of 1545. There were 865 women and 680 men in this dataset, with a median age of 69.8 years (IQR 62.3–75.8). In addition to preoperative data, data were available for three (not necessarily consecutive) annual follow-up assessments in 812 (52.6%), two in 446 (28.9%) and one in 287 patients (18.6%). All individuals exhibited postoperative improvements in SF36/ WOMAC function and pain scores, whilst allowing for the different rates of improvements for each individual (i.e. the random effect of patient ID). However, the rate of improvement was significantly affected by covariates included in the LMEs.
Fig. 1. Predicted annual change in physical function (as measured by SF36), as affected by sex of patient (left = female, right = male), preoperative BMI (x-axis), and change in BMI of at least 10% (coloured groups). Predictions are for individuals at the median age and median initial SF36 scores for pain and physical functioning. For example, a female patient with a preoperative BMI of 30 can expect an annual improvement in SF36-physical function of 20.5 points if she has not gained weight since TKA, but only 16.8 points if she has gained weight.
208
A. Mackie et al. / The Journal of Arthroplasty 30 (2015) 206–209
Fig. 2. Predicted annual change in WOMAC function score, as influenced by preoperative BMI (x-axis) and preoperative WOMAC function score (coloured groups, values predicted are lower quartile, median, and upper quartile of observed data).
− 4.22, P ≤ 0.001) regardless of sex. Improvement in SF36 pain score was negatively related to preoperative SF36 pain score (t = − 13.78, P ≤ 0.001), and positively associated with preoperative SF36 function score (t = 5.55, P ≤ 0.001). The random effect of patient ID accounted for 54.6% of the variance in the model (conditional R 2 = 0.647, marginal R2 = 0.101). Improvement in WOMAC pain score was significantly negatively associated with preoperative BMI (t = −2.64, P ≤ 0.001) and preoperative WOMAC pain score (t = − 21.89, P ≤ 0.001, Fig. 4). The random
Fig. 4. Predicted annual change in WOMAC pain score, as influenced by preoperative BMI (x-axis) and preoperative WOMAC pain score (coloured groups, values predicted are lower quartile, median, and upper quartile of observed data).
effect of patient ID accounted for 57.3% of the variance in the model (conditional R2 = 0.800, marginal R2 = 0.227). Discussion This study has found a relationship between postoperative BMI change following TKA and PROMs. Preoperative BMI was negatively associated with the rate of improvement of all PROMs scores considered. Weight gain of N10% following surgery has an association with lower improvements in SF36 pain and SF36 functional domains. In addition, we identified an inter-relationship between other co-variables (Table 2). Lower preoperative WOMAC pain and function scores were associated with lesser improvement in postoperative WOMAC scores. However, postoperative changes in BMI had no significant effect on postoperative WOMAC pain and function. From this model, age was identified an important determinant of the rate of improvement in SF36 function. Male gender was associated with higher rates of improvement in SF36 function and SF36 pain. Preoperative SF36 pain and SF36 function scores were both significant determinants of improvement in both SF36 pain and SF36 function; however only preoperative WOMAC pain was a significant predictor of the rate of improvement of WOMAC pain, and only preoperative WOMAC Table 2 Regression Coefficients for Each Explanatory Variable.
Fig. 3. Predicted annual change in SF36 pain score, as affected by sex of patient (left = female, right = male), preoperative BMI (x-axis), and change in BMI of at least 10% (coloured groups). Predictions are for individuals at the median age and median initial SF36 scores for pain and functionality.
Explanatory Variable
SF36 Function
SF36 Pain
WOMAC Function
WOMAC Pain
(Intercept) Age Preoperative BMI Sex: male SF36-FUNC preop SF36-PAIN preop WOMAC-FUNC preop WOMAC-PAIN preop 10% change BMI preop: Gain 10% change BMI preop: Loss
67.265 −0.379 −0.533 4.349 −0.511 0.267 * * −3.717 ns
47.249 ns −0.501 3.797 0.202 −0.513 * * −3.818 ns
60.037 ns −0.218 ns * * −0.485 ns ns ns
74.584 ns −0.262 ns * * ns −0.628 ns ns
Values are only provided for those variables that were significant at the 10% level (“ns” indicates not significant). An asterisk (*) indicates that this explanatory variable was not tested for this response variable.
A. Mackie et al. / The Journal of Arthroplasty 30 (2015) 206–209
function was a significant predictor of the rate of improvement of WOMAC function. Owing to the random effect of patient ID being included in these models, all patients had a different intercept for the rate of improvement. For all PROMs the inclusion of preoperative score as a gradient in the random effect improved the model fit (except for SF36 pain, where the algorithm did not converge). In other words, patients with a low preoperative score improved at a different rate to those whose initial score was higher. For SF36 function score, including age-varying gradients (that is, younger patients recovering quicker than older patients) significantly improved model fit. For all four PROMs, 20–33% of the residual variation in the model could be attributed to temporal autocorrelation in the random effect, showing that PROMs recorded in previous postoperative follow-up examinations significantly influenced PROMs recorded at subsequent examinations. Clearly knee arthroplasty is likely to have a greater influence on WOMAC than weight change [14]. However, this study highlights a significant negative effect on both preoperative WOMAC-function and WOMAC-pain scores on postoperative scores respectively. Kaukua et al [15] reported that weight reduction after surgery is associated with improved PROMS score. Unlike our study, they did not demonstrate a gender difference, but theirs was a smaller study (100 patients). They reported that the associated increase in exercise and physical activity can also lead to improvement in obesity-related psychosocial problems; physical functioning; physical role functioning; bodily pain; general health; mental health; and vitality [15]. As all of these factors may impact on the overall SF36 scores of individuals, it may be unsurprising that our study found that weight gain had a negative effect on SF36 scores postoperatively. The use of self-reported height and weight throughout is a limitation of our registry data collection and potentially therefore our study. However, previous studies have demonstrated the accuracy of self-reported height and weight, in the malnourished and overweight/obese populations [16,17]. Secondly this study is based on institutional registry data, which although it includes several surgeons, may not be applicable to the wider population. Institutional registries have many advantages over national registries, including the collection of more detailed data [18]. Furthermore, where there are major geographical differences in practise it may be more appropriate to compare practise within an institution rather than across a whole country. Thirdly, although registry data were collected prospectively the retrospective design of this analysis meant that the data were incomplete – in particular some variables of interest such as complications, length of stay and readmission rates were not available and may have had a significant impact on patient reported outcomes. Finally, we were unable to obtain data about other important factors known to influence patient satisfaction. Factors including mental health status/depression [19,20]; general health status [16]; need for further surgery [19,21] and patient expectations [21] are all known to influence patient satisfaction. As we were unable to measure and adjust for these factors they may be a source of confounding. It is acknowledged that weight maintenance following TKA is met with improved health benefits and improvements in PROMs. 10% increase on preoperative BMIs has significant negative effects on both SF36 function and SF36 pain scores. We believe there is an important clinical message here: patients should avoid weight gain after TKA, as there is a clear association with poorer outcomes. This study has highlighted the importance of weight management following TKA. It drives our focus upon the importance of continued patient education after surgery. There would appear to be an opportunity for an intervention around knee arthroplasty to facilitate this. Previous work has focused upon the role of weight maintenance prior to joint
209
arthroplasty upon functional outcome. Suggested weight loss strategies include dietary support, psychological intervention and more recently bariatric surgery [22–24]. Our work has highlighted the importance of continuing these efforts in the postoperative period. We consider that this work supports the argument for further formal study of a weight management intervention in the postoperative period.
Acknowledgments The authors would like to thank all surgeons contributing patients to the Freeman arthroplasty register (Prof A McCaskie, Mr N Brewster, Mr D Weir, Mr J Holland, Mr M Siddique, Mr A Gray, Mr M Hashmi).
References 1. Gelber AC, Hochberg MC, Mead LA, et al. Body mass index in young men and the risk of subsequent knee and hip osteoarthritis. Am J Med 1999;107:542. 2. Dowsey MM, Choong PF. Early outcomes and complications following joint arthroplasty in obese patients: a review of the published reports. ANZ J Surg 2008; 78:439. 3. Berend ME, Ritter MA, Medling JB, et al. Tibial component failure mechanisms in total knee arthroplasty. Clin Orthop 2004;428:26. 4. Foran JR, Mont MA, Rajadhyaksha AD, et al. Total knee arthroplasty in obese patients: a comparison with a matched control group. J Arthroplasty 2004;19:817. 5. Dowsey MM, Liew D, Stoney JD, et al. The impact of pre-operative obesity on weight change and outcome in total knee replacement. J Bone Joint Surg (Br) 2010;92-B:513. 6. Franz MJ, VanWormer JJ, Crain AL, et al. Weight-loss outcome: a systematic review and meta-analysis of weight-loss clinical trials with a minimum 1-year follow-up. J Am Diet Assoc 2007;107(10):1755. 7. Deshmukh RG, Hayes JH, Pinder IM. Does body weight influence outcome after total knee arthroplasty? A 1-year analysis. J Arthroplasty 2002;17:315. 8. Ware JE, Sherbourne CD. The MOS 36-Item Short-Form health survey (SF36): I. conceptual framework and item selection. Med Care 1992;30:473. 9. Hawker G, Melfi C, Paul J, et al. Comparison of a generic (SF36) and a disease specific (WOMAC) instrument in the measurement of outcomes after knee replacement surgery. J Rheumatol 1995;22:1193. 10. Bellamy N, Buchanan WW, Goldsmith CH, et al. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to anti rheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol 1998;15(12):1833. 11. Pinheiro J, Bates D, DebRoy S, et al. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-104; 2012. 12. R development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2014. 13. Barton K. MuMIn: Multi-model inference. R package version 1.10.0; 2014. 14. Hartley RC, Barton-Hanson NG, Finley R, et al. Early patient outcomes after primary and revision total knee arthroplasty. A prospective study. J Bone Joint Surg (Br) 2002;84(7):994. 15. Kaukua J, Pekkarinen T, Sane T, et al. Health-related quality of life in obese outpatients losing weight with very-low-energy diet and behaviour modification: a 2-y follow-up study. Int J Obes 2003;27:1072. 16. Sangha O, Stucki G, Liang MH, et al. The Self-Administered Comorbidity Questionnaire: a new method to assess co morbidity for clinical and health services research. Arthritis Rheum 2003;49(2):156. 17. Haverkort EB, de Haan RJ, Binnekade JM, et al. Self-reporting of height and weight: valid and reliable identification of malnutrition in preoperative patients. Am J Surg 2012;203(6):700. 18. Blisard Rene. Orthopedics today. http://www.healio.com/orthopedics/total-joint-reconstruction/news/print/orthopedics-today/{7122f0ef-412a-465f-a4cbfee097175025}/national-joint-registry-in-the-us-makes-progress-but-faces-obstacles-in-execution. [date last accessed 30th September 2013]. 19. Stickles B, Phillips L, Brox WT, et al. Defining the relationship between obesity and total joint arthroplasty. Obes Res 2001;9(3):219. 20. Hawker G, Wright J, Coyte P, et al. Health related quality of life after knee replacement. J Bone Joint Surg Am 1998;80(2):163. 21. Dekkers JC, van Wier MF, Hendriksen IJ, et al. Accuracy of self-reported body weight, height and waist circumference in a Dutch overweight working population. BMC Med Res Methodol 2008;28(8):69. 22. Sarwer DB, Moore RH, Spitzer JC, et al. A pilot study investigating the efficacy of postoperative dietary counseling to improve outcomes after bariatric surgery. Surg Obes Relat Dis 2012;8(5):561. 23. Ortega E, Morínigo R, Flores L, et al. Predictive factors of excess body weight loss 1 year after laparoscopic bariatric surgery. Surg Endosc 2012;26(6):1744. 24. Compher CW, Hanlon A, Kang Y, et al. Attendance at clinical visits predicts weight loss after gastric bypass surgery. Obes Surg 2012;22(6):927.
Contents lists available at ScienceDirect
The Journal of Arthroplasty journal homepage: www.arthroplastyjournal.org
Association Between Body Mass Index Change and Outcome in the First Year After Total Knee Arthroplasty Alasdair Mackie, BSc(Hons), MBBS a, Karthikeyan Muthumayandi, MPTh, MPhil a, Mark Shirley, BSc, (Hons), PhD b, David Deehan, MD, MSc, FRCS (T&O) a, Craig Gerrand, MB, ChB, FRCSEd (Tr and Orth), MD a a b
Freeman Hospital, Newcastle upon Tyne, United Kingdom School of Biology, Ridley Building, University of Newcastle, Newcastle upon Tyne, United Kingdom
a r t i c l e
i n f o
Article history: Received 21 May 2014 Accepted 2 September 2014 Keywords: weight change BMI TKA impact knee outcome
a b s t r a c t There is an association between obesity, osteoarthritis and total knee arthroplasty (TKA), but little is known about how postoperative weight change influences outcomes. Primary TKA patients were identified from an institutional arthroplasty registry. BMI and patient reported outcome measures (PROMs, specifically WOMAC and SF36) were recorded for 1545 patients preoperatively and up to 3 years postoperatively. Mixed effects modelling showed postoperative BMI change had no impact on postoperative WOMAC scores. However, weight gain over 10% had a negative impact on SF36 pain and functional scores although postoperative weight loss was not associated with improved PROMs. Men showed greater improvement in postoperative SF36 function and pain scores, whilst older patients were slower to improve. Postoperative weight gain has a negative association with SF36 pain and function. © 2014 Elsevier Inc. All rights reserved.
High body mass index (BMI) is associated with higher prevalence of knee osteoarthritis (OA) and reduced reported mobility [1]. Furthermore, it has been linked to poorer outcomes after total knee arthroplasty (TKA) [2,3] including reduced survival because of aseptic loosening [4]. The crucial barriers to weight loss after surgery remain unclear [5]. Increases in BMI in the general population mean there is an increasing awareness of the need to manage weight and interest in how this might be achieved [6]. When patients undergo knee arthroplasty there is a timely opportunity to address comorbidities including obesity. However, there remains little evidence-based guidance for such patients after knee arthroplasty. Given the recognised associations between BMI, knee arthritis and surgical outcomes, we were interested in exploring the impact of BMI change after surgery on patient reported outcomes in an institutional registry, and what advice we might therefore offer patients. Specifically, we were interested in how postoperative BMI change is related to SF36 and WOMAC scores at a year after surgery. Our null hypothesis was that a significant change in postoperative BMI (arbitrarily taken as ±10% body weight or BMI) would not influence patient reported outcomes.
The Conflict of Interest statement associated with this article can be found at http://dx. doi.org/10.1016/j.arth.2014.09.003. Reprint requests: Alasdair Mackie, BSc, Hons, MBBS, 46 Linden Road, Gosforth, Newcastle-Upon-Tyne, Tyne and Wear, NE3 4HB, United Kingdom. http://dx.doi.org/10.1016/j.arth.2014.09.003 0883-5403/© 2014 Elsevier Inc. All rights reserved.
Patients and Methods The Freeman Joint Registry (FJR) was set up in July 2003 as an ongoing institutional audit of patient outcomes following hip or knee arthroplasty. It is registered with and has approval from the institutional research and development board (Project ID number: 3290). Inclusion in the FJR requires informed consent preoperatively for the collection, storage and analysis of data. The study was conducted in accordance with the declaration of Helsinki and the guidelines for good clinical practise. This study was therefore a retrospective comparative cohort study from a single centre institutional arthroplasty registry using anonymised data. Data items used in this study were preoperative patient demographics including; age, gender, comorbidities and self reported height and weight. Patient reported outcomes included the Western Ontario and McMaster University Osteoarthritis Index (WOMAC), and Medical Outcomes Trust Short Form-36 (SF36) [7–10]. Preoperative assessment was undertaken within 6 weeks of surgery and postoperative analysis was performed annually out to 3 years using both WOMAC and SF36 scores. Years 1 and 2 data are collected in out patients; year 3 data are collected using a postal survey. For the purposes of this analysis, we selected patients who underwent primary TKA over a five year period from April 2004 to March 2009 to allow for the collection of follow up data. All patients underwent cemented TKA using either Press Fit Condylar (PFC) (Depuy, Warsaw, Indiana, USA) or Triathlon (Stryker, Marwah, New Jersey, USA) knee implants. No distinction was made between these
A. Mackie et al. / The Journal of Arthroplasty 30 (2015) 206–209 Table 1 Explanatory Variables Used in the Models. Fixed Effects Sex Age BMI preop SF36-FUNC preop SF36-PAIN preop W-FUNC preop W-PAIN preop 10% change BMI preop
Year of assessment Random Effects ID Age category
Definition Categorical Continuous Preoperative body mass index; continuous Preoperative SF36 physical function; continuous Preoperative SF36 bodily pain; continuous Preoperative WOMAC functionality; continuous Preoperative WOMAC pain; continuous Change in BMI from preoperative assessment, divided into three categories: gain (of ≥10% preoperative BMI), loss (of ≥10% preoperative BMI), no change (b10% gain or loss) Ordinal Definition Anonymised identification code; categorical 30–34.9 35–39.9, …, 90–94.9; ordinal
implants are both are modern generation cruciateretaining minimally constrained. All implants were cemented in place. We used linear mixed effects models (LMEs) for this analysis to investigate the role of key variables (Table 1) in explaining the variation in each of the response variables. LMEs were used because they account for repeated measures on the same individuals by including the unique ID of each patient as a random effect. Each unique ID is fitted in the model with a different gradient; with the aim of reducing the residual error of the model by allowing each individual to improve at his or her own rate, rather than assuming that everyone improves at the same rate. All analyses were performed using the 'nlme' package in the R statistical language [11,12], and pseudo-R-squared values generated using the 'MuMIn' package [13]. Marginal R2 calculated by the latter represents the variance explained by fixed effects, whereas conditional R2 is interpreted as the variance explained by both fixed and random effects. Variance attributable to the random effect is therefore the difference between the two. Four response variables were considered in the model: improvement of function score and improvement of pain score from the preoperative assessment, using two different scales: the Western Ontario and McMaster Universities Arthritis Index (WOMAC) and the Short Form (36) Health Survey (SF36). High WOMAC scores reflect poorer outcomes, whereas higher SF36 scores are associated with better outcomes. The following explanatory variables (Table 1) were identified; age, preoperative BMI, sex, SF36 and WOMAC pain and function, and postoperative weight loss or gain (greater or equal to 10%). Percentage weight change was calculated by the difference of postoperative BMI in relation to the preoperative value for each individual patient.
207
There was no significant increase in the rate of in SF36 function score improvement in patients who lost 10% or more of preoperative BMI over those whose weight did not change postoperatively. However, patients who gained 10% or more of their preoperative BMI following total knee arthroplasty had significantly smaller improvements in physical function (t = −2.67, P = 0.008, Fig. 1). Increasing age and increasing preoperative BMI were associated with slower improvements in the SF36 function score (t = − 5.71, P ≤ 0.001 and t = − 4.25, P ≤ 0.001). Male patients had a significantly larger improvement in SF36 function score than female patients (t = 3.52, P ≤ 0.001). Recovery of SF36 function score was positively related to preoperative assessment of pain (t = 7.02, P ≤ 0.001), and negatively associated with preoperative SF36 function score (t = − 13.83, P ≤ 0.001). Overall, the random effect of patient ID accounted for 58.5% of the variance in the model (conditional R2 = 0.704, marginal R2 = 0.119). Improvement of WOMAC function score was significantly related to preoperative BMI (t = − 2.13, P = 0.033) and preoperative WOMAC function score (t = −15.74, P ≤ 0.001). Both these associations were negative, indicating that patients who had a high preoperative BMI and/or a high preoperative WOMAC function score showed a smaller improvement in WOMAC function score than patients with lower preoperative BMI or WOMAC function scores (Fig. 2). The random effect of patient ID accounted for 64.9% of the variance in the model (conditional R 2 = 0.782, marginal R2 = 0.133). Individuals who gained more than 10% of their preoperative BMI following a total knee arthroplasty showed a significant reduction in the rate of improvement of SF36 pain score (t = − 2.58, P = 0.01) compared to those whose weight remained the same or decreased. Weight gain of less than 10% of preoperative BMI was not associated with less improvement in SF36 pain (Fig. 3). Unlike improvements in SF36 function score, age was not a significant predictor of improvement in SF36 pain score. Male patients had a significantly larger improvement in pain score than female patients (t = 3.20, P = 0.001). Preoperative BMI was negatively associated with the rate of improvement (t =
Results There were 1902 patients in the original dataset, however, after accounting for missing data on variables of interest (age, sex, BMI, PROMS for pain and functionality) there were records for 1821 individuals. Of these, 276 without follow up data were excluded, leaving a sample size of 1545. There were 865 women and 680 men in this dataset, with a median age of 69.8 years (IQR 62.3–75.8). In addition to preoperative data, data were available for three (not necessarily consecutive) annual follow-up assessments in 812 (52.6%), two in 446 (28.9%) and one in 287 patients (18.6%). All individuals exhibited postoperative improvements in SF36/ WOMAC function and pain scores, whilst allowing for the different rates of improvements for each individual (i.e. the random effect of patient ID). However, the rate of improvement was significantly affected by covariates included in the LMEs.
Fig. 1. Predicted annual change in physical function (as measured by SF36), as affected by sex of patient (left = female, right = male), preoperative BMI (x-axis), and change in BMI of at least 10% (coloured groups). Predictions are for individuals at the median age and median initial SF36 scores for pain and physical functioning. For example, a female patient with a preoperative BMI of 30 can expect an annual improvement in SF36-physical function of 20.5 points if she has not gained weight since TKA, but only 16.8 points if she has gained weight.
208
A. Mackie et al. / The Journal of Arthroplasty 30 (2015) 206–209
Fig. 2. Predicted annual change in WOMAC function score, as influenced by preoperative BMI (x-axis) and preoperative WOMAC function score (coloured groups, values predicted are lower quartile, median, and upper quartile of observed data).
− 4.22, P ≤ 0.001) regardless of sex. Improvement in SF36 pain score was negatively related to preoperative SF36 pain score (t = − 13.78, P ≤ 0.001), and positively associated with preoperative SF36 function score (t = 5.55, P ≤ 0.001). The random effect of patient ID accounted for 54.6% of the variance in the model (conditional R 2 = 0.647, marginal R2 = 0.101). Improvement in WOMAC pain score was significantly negatively associated with preoperative BMI (t = −2.64, P ≤ 0.001) and preoperative WOMAC pain score (t = − 21.89, P ≤ 0.001, Fig. 4). The random
Fig. 4. Predicted annual change in WOMAC pain score, as influenced by preoperative BMI (x-axis) and preoperative WOMAC pain score (coloured groups, values predicted are lower quartile, median, and upper quartile of observed data).
effect of patient ID accounted for 57.3% of the variance in the model (conditional R2 = 0.800, marginal R2 = 0.227). Discussion This study has found a relationship between postoperative BMI change following TKA and PROMs. Preoperative BMI was negatively associated with the rate of improvement of all PROMs scores considered. Weight gain of N10% following surgery has an association with lower improvements in SF36 pain and SF36 functional domains. In addition, we identified an inter-relationship between other co-variables (Table 2). Lower preoperative WOMAC pain and function scores were associated with lesser improvement in postoperative WOMAC scores. However, postoperative changes in BMI had no significant effect on postoperative WOMAC pain and function. From this model, age was identified an important determinant of the rate of improvement in SF36 function. Male gender was associated with higher rates of improvement in SF36 function and SF36 pain. Preoperative SF36 pain and SF36 function scores were both significant determinants of improvement in both SF36 pain and SF36 function; however only preoperative WOMAC pain was a significant predictor of the rate of improvement of WOMAC pain, and only preoperative WOMAC Table 2 Regression Coefficients for Each Explanatory Variable.
Fig. 3. Predicted annual change in SF36 pain score, as affected by sex of patient (left = female, right = male), preoperative BMI (x-axis), and change in BMI of at least 10% (coloured groups). Predictions are for individuals at the median age and median initial SF36 scores for pain and functionality.
Explanatory Variable
SF36 Function
SF36 Pain
WOMAC Function
WOMAC Pain
(Intercept) Age Preoperative BMI Sex: male SF36-FUNC preop SF36-PAIN preop WOMAC-FUNC preop WOMAC-PAIN preop 10% change BMI preop: Gain 10% change BMI preop: Loss
67.265 −0.379 −0.533 4.349 −0.511 0.267 * * −3.717 ns
47.249 ns −0.501 3.797 0.202 −0.513 * * −3.818 ns
60.037 ns −0.218 ns * * −0.485 ns ns ns
74.584 ns −0.262 ns * * ns −0.628 ns ns
Values are only provided for those variables that were significant at the 10% level (“ns” indicates not significant). An asterisk (*) indicates that this explanatory variable was not tested for this response variable.
A. Mackie et al. / The Journal of Arthroplasty 30 (2015) 206–209
function was a significant predictor of the rate of improvement of WOMAC function. Owing to the random effect of patient ID being included in these models, all patients had a different intercept for the rate of improvement. For all PROMs the inclusion of preoperative score as a gradient in the random effect improved the model fit (except for SF36 pain, where the algorithm did not converge). In other words, patients with a low preoperative score improved at a different rate to those whose initial score was higher. For SF36 function score, including age-varying gradients (that is, younger patients recovering quicker than older patients) significantly improved model fit. For all four PROMs, 20–33% of the residual variation in the model could be attributed to temporal autocorrelation in the random effect, showing that PROMs recorded in previous postoperative follow-up examinations significantly influenced PROMs recorded at subsequent examinations. Clearly knee arthroplasty is likely to have a greater influence on WOMAC than weight change [14]. However, this study highlights a significant negative effect on both preoperative WOMAC-function and WOMAC-pain scores on postoperative scores respectively. Kaukua et al [15] reported that weight reduction after surgery is associated with improved PROMS score. Unlike our study, they did not demonstrate a gender difference, but theirs was a smaller study (100 patients). They reported that the associated increase in exercise and physical activity can also lead to improvement in obesity-related psychosocial problems; physical functioning; physical role functioning; bodily pain; general health; mental health; and vitality [15]. As all of these factors may impact on the overall SF36 scores of individuals, it may be unsurprising that our study found that weight gain had a negative effect on SF36 scores postoperatively. The use of self-reported height and weight throughout is a limitation of our registry data collection and potentially therefore our study. However, previous studies have demonstrated the accuracy of self-reported height and weight, in the malnourished and overweight/obese populations [16,17]. Secondly this study is based on institutional registry data, which although it includes several surgeons, may not be applicable to the wider population. Institutional registries have many advantages over national registries, including the collection of more detailed data [18]. Furthermore, where there are major geographical differences in practise it may be more appropriate to compare practise within an institution rather than across a whole country. Thirdly, although registry data were collected prospectively the retrospective design of this analysis meant that the data were incomplete – in particular some variables of interest such as complications, length of stay and readmission rates were not available and may have had a significant impact on patient reported outcomes. Finally, we were unable to obtain data about other important factors known to influence patient satisfaction. Factors including mental health status/depression [19,20]; general health status [16]; need for further surgery [19,21] and patient expectations [21] are all known to influence patient satisfaction. As we were unable to measure and adjust for these factors they may be a source of confounding. It is acknowledged that weight maintenance following TKA is met with improved health benefits and improvements in PROMs. 10% increase on preoperative BMIs has significant negative effects on both SF36 function and SF36 pain scores. We believe there is an important clinical message here: patients should avoid weight gain after TKA, as there is a clear association with poorer outcomes. This study has highlighted the importance of weight management following TKA. It drives our focus upon the importance of continued patient education after surgery. There would appear to be an opportunity for an intervention around knee arthroplasty to facilitate this. Previous work has focused upon the role of weight maintenance prior to joint
209
arthroplasty upon functional outcome. Suggested weight loss strategies include dietary support, psychological intervention and more recently bariatric surgery [22–24]. Our work has highlighted the importance of continuing these efforts in the postoperative period. We consider that this work supports the argument for further formal study of a weight management intervention in the postoperative period.
Acknowledgments The authors would like to thank all surgeons contributing patients to the Freeman arthroplasty register (Prof A McCaskie, Mr N Brewster, Mr D Weir, Mr J Holland, Mr M Siddique, Mr A Gray, Mr M Hashmi).
References 1. Gelber AC, Hochberg MC, Mead LA, et al. Body mass index in young men and the risk of subsequent knee and hip osteoarthritis. Am J Med 1999;107:542. 2. Dowsey MM, Choong PF. Early outcomes and complications following joint arthroplasty in obese patients: a review of the published reports. ANZ J Surg 2008; 78:439. 3. Berend ME, Ritter MA, Medling JB, et al. Tibial component failure mechanisms in total knee arthroplasty. Clin Orthop 2004;428:26. 4. Foran JR, Mont MA, Rajadhyaksha AD, et al. Total knee arthroplasty in obese patients: a comparison with a matched control group. J Arthroplasty 2004;19:817. 5. Dowsey MM, Liew D, Stoney JD, et al. The impact of pre-operative obesity on weight change and outcome in total knee replacement. J Bone Joint Surg (Br) 2010;92-B:513. 6. Franz MJ, VanWormer JJ, Crain AL, et al. Weight-loss outcome: a systematic review and meta-analysis of weight-loss clinical trials with a minimum 1-year follow-up. J Am Diet Assoc 2007;107(10):1755. 7. Deshmukh RG, Hayes JH, Pinder IM. Does body weight influence outcome after total knee arthroplasty? A 1-year analysis. J Arthroplasty 2002;17:315. 8. Ware JE, Sherbourne CD. The MOS 36-Item Short-Form health survey (SF36): I. conceptual framework and item selection. Med Care 1992;30:473. 9. Hawker G, Melfi C, Paul J, et al. Comparison of a generic (SF36) and a disease specific (WOMAC) instrument in the measurement of outcomes after knee replacement surgery. J Rheumatol 1995;22:1193. 10. Bellamy N, Buchanan WW, Goldsmith CH, et al. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to anti rheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol 1998;15(12):1833. 11. Pinheiro J, Bates D, DebRoy S, et al. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-104; 2012. 12. R development Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2014. 13. Barton K. MuMIn: Multi-model inference. R package version 1.10.0; 2014. 14. Hartley RC, Barton-Hanson NG, Finley R, et al. Early patient outcomes after primary and revision total knee arthroplasty. A prospective study. J Bone Joint Surg (Br) 2002;84(7):994. 15. Kaukua J, Pekkarinen T, Sane T, et al. Health-related quality of life in obese outpatients losing weight with very-low-energy diet and behaviour modification: a 2-y follow-up study. Int J Obes 2003;27:1072. 16. Sangha O, Stucki G, Liang MH, et al. The Self-Administered Comorbidity Questionnaire: a new method to assess co morbidity for clinical and health services research. Arthritis Rheum 2003;49(2):156. 17. Haverkort EB, de Haan RJ, Binnekade JM, et al. Self-reporting of height and weight: valid and reliable identification of malnutrition in preoperative patients. Am J Surg 2012;203(6):700. 18. Blisard Rene. Orthopedics today. http://www.healio.com/orthopedics/total-joint-reconstruction/news/print/orthopedics-today/{7122f0ef-412a-465f-a4cbfee097175025}/national-joint-registry-in-the-us-makes-progress-but-faces-obstacles-in-execution. [date last accessed 30th September 2013]. 19. Stickles B, Phillips L, Brox WT, et al. Defining the relationship between obesity and total joint arthroplasty. Obes Res 2001;9(3):219. 20. Hawker G, Wright J, Coyte P, et al. Health related quality of life after knee replacement. J Bone Joint Surg Am 1998;80(2):163. 21. Dekkers JC, van Wier MF, Hendriksen IJ, et al. Accuracy of self-reported body weight, height and waist circumference in a Dutch overweight working population. BMC Med Res Methodol 2008;28(8):69. 22. Sarwer DB, Moore RH, Spitzer JC, et al. A pilot study investigating the efficacy of postoperative dietary counseling to improve outcomes after bariatric surgery. Surg Obes Relat Dis 2012;8(5):561. 23. Ortega E, Morínigo R, Flores L, et al. Predictive factors of excess body weight loss 1 year after laparoscopic bariatric surgery. Surg Endosc 2012;26(6):1744. 24. Compher CW, Hanlon A, Kang Y, et al. Attendance at clinical visits predicts weight loss after gastric bypass surgery. Obes Surg 2012;22(6):927.
