Clinically Important Body Weight Gain Following Knee Arthroplasty: A Five-Year Comparative Cohort Study

Authors

  • Daniel L. Riddle,

    Corresponding author
    1. Virginia Commonwealth University, Richmond
    • Virginia Commonwealth University, Departments of Physical Therapy and Orthopaedic Surgery, PO Box 980224, Richmond, VA 23298-0224

    Search for more papers by this author
    • Dr. Riddle has received honoraria (less than $10,000) for his role as Deputy Editor of Physical Therapy.

  • Jasvinder A. Singh,

    1. Birmingham Veterans Affairs Medical Center and University of Alabama at Birmingham
    Search for more papers by this author
    • Dr. Singh has received consultant fees, speaking fees, and/or honoraria (less than $10,000 each) from URL, Savient, and Novartis, and (more than $10,000 each) from Takeda, Allergan, and Ardea, and has received investigator-initiated grants from Takeda and Savient.

  • William S. Harmsen,

    1. Mayo Clinic, Rochester, Minnesota
    Search for more papers by this author
  • Cathy D. Schleck,

    1. Mayo Clinic, Rochester, Minnesota
    Search for more papers by this author
  • David G. Lewallen

    1. Mayo Clinic, Rochester, Minnesota
    Search for more papers by this author
    • Dr. Lewallen has received consultant fees, speaking fees, and/or honoraria (less than $10,000 each) from Osteotech and Zimmer, and receives fees and royalties from Zimmer for hip and knee implants.


Abstract

Objective

The impact of knee arthroplasty on body weight has not been fully explored. Clinically important weight gain following knee arthroplasty would pose potentially important health risks.

Methods

We used one of the largest US-based knee arthroplasty registries and a population-based control sample from the same geographic region to determine whether knee arthroplasty increases the risk of clinically important weight gain of ≥5% of baseline body weight over a 5-year postoperative period.

Results

Of the persons in the knee arthroplasty sample, 30.0% gained ≥5% of baseline body weight 5 years following surgery as compared to 19.7% of the control sample. The multivariable-adjusted (age, sex, body mass index, education, comorbidity, and presurgical weight change) odds ratio (OR) was 1.6 (95% confidence interval [95% CI] 1.2–2.2) in persons with knee arthroplasty as compared to the control sample. Additional arthroplasty procedures during followup further increased the risk for weight gain (OR 2.1, 95% CI 1.4–3.1) relative to the control sample. Specifically, among patients with knee arthroplasty, younger patients and those who lost greater amounts of weight in the 5-year preoperative period were at greater risk for clinically important weight gain.

Conclusion

Patients who undergo knee arthroplasty are at an increased risk of clinically important weight gain following surgery. The findings potentially have broad implications to multiple members of the health care team. Future research should develop weight loss/maintenance interventions particularly for younger patients who have lost a substantial amount of weight prior to surgery, as they are most at risk for substantial postsurgical weight gain.

INTRODUCTION

Managing patients who are overweight or obese is one of the major challenges facing orthopedic surgeons caring for patients with knee osteoarthritis (OA). Obesity increases the risk of knee OA by 4–5-fold (1) and is especially high (2), consequential (3, 4), and costly (5) for persons with lower extremity arthroplasty (2, 6). Total knee arthroplasty (TKA) is the most common type of arthroplasty surgery, with more than 700,000 procedures conducted in 2008 (7) and a projected 3.5 million in 2030 (8). Given the well-established link between obesity in older adults and the increased risk for cardiovascular disease and some cancers (9), many patients who undergo TKA are at a particularly high risk for adverse health events as a consequence of excessive weight and, potentially, weight gain after surgery (10). Public health implications of weight gain are particularly important for persons in their 50s and 60s, which is a common age range for TKA. Weight gain during these decades is more likely than in older age groups, which contributes to a greater risk of obesity-related disorders (11–13).

The logical assumption may be that persons who are overweight or obese prior to surgery are more likely to lose weight following surgery. Because there is less pain and improved mobility, the impediments to increased activity and exercise are eased following surgery, and weight loss would logically follow. However, research exploring the impact of TKA on subsequent weight change has been inconclusive. One study reported that a majority of patients gain weight after surgery (10), while other studies report essentially no change (14–18) or overall weight loss (19). Variations among this evidence may be due to a variety of issues. Some studies, for example, examined small sample sizes and may have been underpowered (14, 17). Followup times varied from 3 months to 2 years, and no studies recruited a population-based sample of similarly aged persons without TKA to account for weight gain attributable to normal aging (20–22). In the only study reporting significant weight gains, patients gained an average of 2.5 kg over the 2 years, while the convenience sample of 31 control subjects gained 0.2 kg (10).

We found no literature that compared longer-term weight changes in a large cohort of patients with TKA to a population-based control sample. The purpose of our study was to determine if persons undergoing TKA are at a greater risk for clinically important weight gain (defined as a 5% or greater increase in weight relative to baseline) (23–26) over a 5-year period as compared to an age- and sex-matched population-based sample. Our second purpose was to identify variables that increased the risk of clinically important weight gain specifically among persons with TKA.

Significance & Innovations

  • Relative to a population-based sample, persons undergoing knee arthroplasty are at risk for clinically important 5-year postoperative weight gain.

  • Additional lower extremity arthroplasty procedures increase the risk of clinically important weight gain relative to a population-based sample.

  • Subgroups of patients undergoing knee arthroplasty who are ages <70 years and who have lost a substantial amount of weight prior to surgery are at a particularly high risk for postoperative weight gain.

SUBJECTS AND METHODS

Study samples.

We used 2 data sources. The first was the Mayo Clinic electronic and administrative knee joint registry databases. The registry prospectively collects followup outcomes data at 2- and 5-year intervals from all persons who have provided informed consent and have undergone TKA at Mayo Clinic. The registry has been functional since 1971.

All primary (not revision) TKA surgeries during the period of January 1, 1995 to December 31, 2005 were considered for study inclusion. This time period allowed us to 1) collect data on a large sample of patients receiving TKA, 2) use consistent definitions of both potential confounders and similar methods of height and weight data collection during clinic visits, and 3) potentially collect 5-year followup data on all persons. We excluded persons who had the following medical procedures or disorders ≤2 years prior to or up to 5 years following the index arthroplasty date: bariatric surgery (n = 17) and all forms of cancer other than basal cell carcinoma because of the risk of cancer-related weight loss (n = 135). In addition, 25 persons undergoing arthroplasty did not provide informed consent and 60 persons underwent bilateral TKA on the same day and were excluded. The American Society of Anesthesiologists score (27, 28) was retrieved using a registry maintained by the Department of Anesthesiology. There were 1,122 TKAs in 917 patients, meeting the criteria for inclusion in the study over the time period. Overall, yearly body weight collection rates from the chart reviews for the 5 years following surgery ranged from 99.2% at baseline to 64.3% at 5 years.

The second data source was the Rochester Epidemiology Project (REP), a population-based medical records linkage system based in Olmsted County, Minnesota, home of the Mayo Clinic. The REP has been continuously funded by the National Institutes of Health since 1966 (29). Previous work has indicated that in any 3-year period, more than 90% of Olmsted County residents are examined in the Mayo Clinic health care system, which includes the Mayo Clinic, Mayo Clinic-St. Mary's Hospital, and Mayo Clinic-Methodist Hospital in Rochester, Minnesota (30). The REP has been used extensively to study the epidemiology of a variety of health problems, including rheumatic diseases (31) and obesity (32). The Institutional Review Board at the Mayo Clinic approved the study and all patients signed consent to have their data included in either the knee arthroplasty or the REP databases.

To estimate weight changes over time in the general population, we used the REP to select from all persons residing in Olmsted County and matched, within 30 days, to the dates of the TKA surgeries based on clinic visit dates. We selected a random sample of 237 persons using frequency matching and stratifying on age (groupings of 18–55, 56–65, 66–75, and >75 years), sex, and year of TKA (1995–2005) from the approximately 124,000 people (2000 US Census) in Olmsted County. Persons with a diagnosis of any form of cancer other than basal cell carcinoma or who underwent bariatric surgery during the study period were excluded. For the REP data, body weight collection rates ranged from 100% for the baseline surgery index date to 56% for the 5-year postoperative index date.

A total of 3 trained chart abstractors extracted data from the charts of all persons in both the TKA cohort and the random sample of REP community residents for a total of 1,154 chart reviews. All weight data used in the study were collected from validated weight scales in the Mayo Clinic system. Abstractors reviewed the medical charts for height, weight, and education status. Deyo-Charlson comorbidity scores (33, 34) and cancer diagnoses were found through electronic medical record searches. International Classification of Diseases, Ninth Revision diagnosis codes, or cancer diagnostic labels in cases where abstractors found additional cancer diagnoses, indicated the presence of all cancers, not including basal cell cancer, during the study period (1995–2005).

An a priori power assessment was conducted to estimate the number of persons required to detect a difference among the TKA and REP groups. Assuming a common SD for weight change over 5 years of 4.5 kg (10, 16) and using a 2-sided, 2-sample t-test to compare 5-year weight change, we needed approximately 1,100 patients with TKA and 235 community residents to provide greater than 90% power to detect a difference in group means of 1 kg or more.

Assessment of body weight and body mass index (BMI).

During each medical visit, whether for TKA followup or for any type of medical visit, all persons in the Mayo Clinic health system undergo routine height and weight measurements on digital scales that are calibrated yearly by hospital and clinic staff. For the patients undergoing TKA, height and weight data were collected on the day of surgery (i.e., the index visit). Chart abstractors extracted yearly height and weight data each year for 5 years prior to and 5 years after the TKA surgery date (in total, this covered the years 1990–2010, depending on the index surgery date). Abstractors were instructed to extract yearly height and weight data that were as close to the yearly anniversary dates of the index surgery date as possible. For the REP subjects, chart abstractors extracted height and weight data 5 years prior to and 5 years following the matched cases' TKA index dates. The flow of patients through the study for both the REP and TKA samples is shown in Figure 1.

Figure 1.

The flow of knee arthroplasty and Rochester Epidemiology Project patients through the study and the number of patients with body weight data at each time point. KA = knee arthroplasty.

Outcome variable of interest.

All subjects were categorized as weight gainers if they gained ≥5% of body weight from the index date to the final followup. Persons who did not gain ≥5% of their body weight from the index date to the final followup were classified as weight maintainers. A ≥5% gain in body weight has been found in multiple studies to lead to clinically meaningful effects on cardiovascular and diabetes mellitus–related risk as well as pain and function (9, 25, 26), and has been recommended as a threshold for clinically meaningful weight loss in multiple guidelines (23, 24).

Potential baseline confounder variables.

We examined the following variables in the TKA and REP control cohorts in both univariate and multivariable analyses: 1) age, categorized in approximate quartiles as follows: <60 years, 60–69 years, 70–75 years, and >75 years; 2) sex; 3) comorbidity, assessed using the Deyo-Charlson Index, a validated comorbidity scale that includes 17 comorbidities, with higher scores indicating greater comorbidity (33, 34); 4) baseline BMI (kg/m2), categorized into <25, 25.0 to <30, 30 to <35, and ≥35, with <25 (mild thinness to normal weight) being the referent group (35); 5) education, categorized into 3 groups (less than high school diploma, high school diploma or some college, and at least 4 years of college); 6) contralateral or hip arthroplasty procedures conducted during the 5-year postoperative period; and 7) 2 additional variables created to account for the extent of weight gain in the 5 years prior to surgery. For maximum pre-TKA weight loss, we subtracted the largest weight in the 5 years prior to surgery from the index weight. If the subject demonstrated no weight loss in the 5 years prior to the index date, this variable was coded as 0. A similar strategy was used to indicate the extent of weight gain up to 5 years prior to the index surgery date. We reasoned that if the subject had either lost or gained weight in the years leading up to the index date, the extent of weight gain or loss following surgery may be impacted by this presurgical weight variation (36).

Statistical analyses.

Descriptive statistics are reported for the characteristics of the TKA patients and the REP community residents. A t-test or chi-square test, as appropriate, was used to compare the baseline characteristics of patients with those of the population-based controls (Table 1) and those with and those without missing data at years 2 and 4 (Table 2).

Table 1. Baseline characteristics and extent of missing followup weight data for the study samples*
 REP controls (n = 237)TKA cases (n = 917)P
  • *

    Values are the mean ± SD for continuous variables and the number (percentage) for categorical variables unless otherwise indicated. REP = Rochester Epidemiology Project; TKA = total knee arthroplasty; HS = high school; BMI = body mass index.

  • P value for t-test was used for continuous variables and for chi-square test for categorical variables.

  • Maximum weight loss (in kg), relative to baseline, during the 5 years prior to the baseline knee arthroplasty.

  • §

    Maximum weight gain (in kg), relative to baseline, during the 5 years prior to the baseline knee arthroplasty.

Baseline weight, kg76.3 ± 17.189.1 ± 20.9< 0.001
 Range29–12542–184 
Male sex85 (36)331 (36)0.96
Education  0.13
 Less than HS24 (11)156 (14) 
 HS or some college142 (68)659 (60) 
 ≥4 years of college43 (21)275 (25) 
Deyo-Charlson Index  0.48
 090 (38)365 (40) 
 145 (19)196 (21) 
 >1102 (43)357 (39) 
Age, years  0.67
 <5018 (8)54 (6) 
 50–5935 (15)161 (18) 
 60–6973 (31)298 (32) 
 70–8083 (35)310 (34) 
 >8028 (12)95 (10) 
Baseline BMI, kg/m2  < 0.001
 <2580 (34)124 (14) 
 25–29.987 (37)273 (30) 
 30–34.945 (19)246 (27) 
 35–39.917 (7)160 (18) 
 ≥408 (3)108 (12) 
1-year data present185 (78)628 (68)0.003
2-year data present167 (70)647 (70)0.93
3-year data present171 (72)635 (69)0.47
4-year data present169 (71)640 (70)0.85
5-year data present132 (56)590 (64)0.85
Presurgical 5-year max weight loss, kg3.2 ± 4.04.5 ± 6.00.001
Presurgical 5-year max weight gain, kg§3.2 ± 5.03.7 ± 5.40.20
Table 2. Relationship between nonmissing 2- and 4-year weight data and baseline characteristics for the total sample*
 Nonmissing 2-year dataNonmissing 4-year data
No. (%)PNo. (%)P
  • *

    HS = high school; BMI = body mass index.

  • P value for chi-square test for categorical variables.

Sex 0.41 0.26
 Female527 (71) 526 (71) 
 Male287 (69) 283 (68) 
Education 0.16 0.14
 Less than HS98 (65) 97 (65) 
 HS or some college498 (73) 495 (73) 
 ≥4 years of college190 (70) 196 (73) 
Deyo-Charlson Index < 0.001 0.007
 0294 (65) 315 (69) 
 1192 (80) 188 (78) 
 >1328 (71) 306 (67) 
Age, years 0.49 0.27
 <5059 (68) 50 (69) 
 50–59135 (69) 132 (67) 
 60–69263 (71) 272 (73) 
 70–80287 (73) 277 (70) 
 >8080 (65) 78 (63) 
Baseline BMI, kg/m2 0.88 0.08
 <25144 (71) 140 (69) 
 25–29.9259 (72) 241 (67) 
 30–34.9210 (72) 223 (77) 
 35–39.9122 (69) 121 (68) 
 ≥4079 (68) 84 (72) 

Logistic regression was used to analyze variables associated with clinically important weight gain using all data collected over the 5-year postoperative period and accounting for within-subject correlation (Proc Genmod in SAS with repeated statement). Clinically important weight gain was defined as a ≥5% body weight gain during the 5-year followup as compared to baseline. All weight/height assessments postsurgery were included in the analyses. In addition to subject group (TKA versus REP), other patient factors considered in these models included baseline BMI, age, sex, education, Deyo-Charlson Index, maximum loss in weight from 5 years prior to the index date, and maximum gain in weight 5 years prior to the index date. In all analyses, we examined for all two-way interactions with other variables in the models. The alpha level was set at 0.05 for statistical significance. We conducted a sensitivity analysis to account for the imbalance in baseline BMI distributions between the TKA and REP groups. The marginal structural models method as described by Robins and colleagues (37) uses case weights to balance the distribution of baseline BMI between the TKA and REP patients to allow for statistical comparisons when baseline BMI distribution was equivalent.

A separate logistic regression analysis was conducted to identify variables associated with clinically important weight gain, specifically in the TKA group. For this analysis, we also included as predictor variables diagnosis (OA versus rheumatoid arthritis) and the American Society of Anesthesiologists score, a validated measure of immediate postoperative morbidity and perioperative mortality, and scored as classes I or II versus III or IV (27, 28). Statistical analyses were performed with SAS statistical software, version 9.2.

RESULTS

The TKA sample (n = 917 persons with 205 undergoing additional primary hip or knee arthroplasty procedures during the 5-year followup) and the population-based control sample descriptions are shown in Table 1. A total of 98.3% of the TKA sample and 93.8% of the REP sample contributed followup weight data to the analysis, with 78.2% of the TKA sample and 80.6% of the REP sample with at least 3 years' worth of followup weight data. Most attributes for both groups were similar, with the exception of baseline BMI and weight, which were significantly higher for the TKA sample. With the exception of Deyo-Charlson comorbidity scores, characteristics of the total sample with and without missing data at years 2 and 4 were similar (Table 2).

Over the 5-year postoperative period, the control group (n = 237) lost a mean ± SD of −0.35 ± 5.0 kg. The patients with subsequent additional knee or hip arthroplasty(ies) during the 5-year followup gained a mean ± SD of 2.62 ± 6.0 kg, while cases with no additional lower extremity arthroplasty during 5 years of followup gained a mean ± SD of 1.23 ± 6.4 kg. Percentages of patients in the TKA sample who gained ≥5% of body weight varied from 22.1% in year 1 to 32.3% in year 4, whereas 16.0% to 22.8% of the REP sample gained ≥5% of body weight over the study period (Table 3).

Table 3. Percentage of subjects in the total knee arthroplasty and Rochester Epidemiology Project samples who gained a clinically important amount of body weight over the 5-year followup period
YearTotal knee arthroplastyRochester Epidemiology Project
NNo. (%) with clinically important gainNNo. (%) with clinically important gain
1624138 (22.1)18530 (16.2)
2647183 (28.3)16738 (22.8)
3635196 (30.9)17137 (21.6)
4640207 (32.3)16927 (16.0)
5590177 (30.0)13226 (19.7)

In a univariate analysis, patients undergoing a single TKA had odds of gaining ≥5% of body weight up to 5 years following surgery that were 70% higher than those of the REP control group (odds ratio [OR] 1.7, 95% confidence interval [95% CI] 1.3–2.3). Patients undergoing at least 1 additional arthroplasty during the followup period had increased odds (OR 2.3, 95% CI 1.6–3.2) of clinically important weight gain relative to the REP controls. In a multivariable model that adjusted for potential confounders, estimates for clinically important weight gain in persons with a single TKA and persons with additional arthroplasty surgery were similar to those found in the univariate models (Table 4). In the sensitivity analysis that adjusted for the imbalance in baseline BMI between the REP and TKA samples, the key estimates were very similar to those reported in Supplementary Appendix A (available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.21880/abstract).

Table 4. Logistic regression model of clinically important weight gain (≥5% of body weight) and TKA case/REP control status*
 Unadjusted analysisAdjusted analysis
OR (95% CI)POR (95% CI)P
  • *

    TKA = total knee arthroplasty; REP = Rochester Epidemiology Project; OR = odds ratio; 95% CI = 95% confidence interval; BMI = body mass index; HS = high school.

  • These ORs were significant in the analyses.

  • Maximum weight loss (in kg), relative to baseline, during the 5 years prior to the baseline knee arthroplasty.

  • §

    Maximum weight gain (in kg), relative to baseline, during the 5 years prior to the baseline knee arthroplasty.

Subject status    
 TKA (followup arthroplasty)2.3 (1.6–3.2)< 0.0012.1 (1.4–3.1)< 0.001
 TKA (no followup arthroplasty)1.7 (1.3–2.3)< 0.0011.6 (1.2–2.2)0.003
 REP control1.0 (reference) 1.0 (reference) 
BMI, kg/m2    
 ≥25.0 to <30.00.7 (0.5–1.0)0.0540.7 (0.5–1.0)0.09
 ≥30.0 to <35.00.9 (0.7–1.3)0.580.8 (0.6–1.2)0.24
 ≥350.8 (0.6–1.1)0.200.6 (0.4–0.9)0.02
 <251.0 (reference) 1.0 (reference) 
Age, years    
 <602.1 (1.5–2.8)< 0.0012.7 (1.8–3.9)< 0.001
 60–691.5 (1.1–2.0)0.011.7 (1.2–2.3)0.004
 70–751.1 (0.9–1.4)0.551.2 (0.8–1.8)0.36
 >751.0 (reference) 1.0 (reference) 
Sex    
 Male1.1 (0.9–1.4)0.411.2 (0.9–1.5)0.19
 Female1.0 (reference) 1.0 (reference) 
Education    
 Less than HS diploma1.0 (0.7–1.5)0.921.2 (0.8–1.7)0.50
 HS or some college1.0 (0.7–1.2)0.701.0 (0.7–1.3)0.86
 College degree or more1.0 (reference) 1.0 (reference) 
Deyo-Charlson Index    
 00.9 (0.7–1.1)0.190.8 (0.6–1.0)0.08
 11.0 (0.8–1.4)0.890.9 (0.6–1.2)0.36
 >11.0 (reference) 1.0 (reference) 
Presurgical 5-year max weight loss, per 1 kg1.10 (1.07–1.13)< 0.0011.10 (1.07–1.13)< 0.001
Presurgical 5-year max weight gain, per 1 kg§0.99 (0.97–1.01)0.920.99 (0.97–1.01)0.44

For the multivariable logistic regression model examining only patients in the TKA group, those who were at higher risk for clinically important weight gain were ages <70 years and had lost more weight in the 5 years prior to surgery. Baseline BMI was not related to risk of ≥5% weight gain in the TKA group (Table 5).

Table 5. Logistic regression model of clinically important weight gain (≥5% of body weight) among persons with knee arthroplasty*
 Unadjusted analysisAdjusted analysis
OR (95% CI)POR (95% CI)P
  • *

    OR = odds ratio; 95% CI = 95% confidence interval; TKA = total knee arthroplasty; BMI = body mass index; HS = high school; ASA = American Society of Anesthesiologists.

  • These ORs were significant in the analyses.

  • Maximum weight loss (in kg), relative to baseline, during the 5 years prior to the baseline knee arthroplasty.

  • §

    Maximum weight gain (in kg), relative to baseline, during the 5 years prior to the baseline knee arthroplasty.

Subject status    
 TKA case (followup arthroplasty)1.3 (1.0–1.7)0.021.3 (0.9–1.7)0.13
 TKA case (no followup arthroplasty)1.0 (reference) 1.0 (reference) 
BMI, kg/m2    
 ≥25.0 to <30.00.7 (0.5–1.0)0.080.8 (0.5–1.2)0.25
 ≥30.0 to <35.00.8 (0.6–1.2)0.350.8 (0.5–1.3)0.36
 ≥350.8 (0.5–1.1)0.130.8 (0.5–1.2)0.23
 <251.0 (reference) 1.0 (reference) 
Age, years    
 <602.2 (1.6–3.1)< 0.0013.0 (1.9–4.8)< 0.001
 60–691.5 (1.1–2.0)0.021.6 (1.1–2.4)0.02
 70–751.2 (0.8–1.7)0.341.4 (0.9–2.2)0.13
 >751.0 (reference) 1.0 (reference) 
Sex    
 Male1.1 (0.8–1.4)0.651.2 (0.9–1.5)0.34
 Female1.0 (reference) 1.0 (reference) 
Diagnosis    
 Osteoarthritis0.8 (0.5–1.1)0.131.0 (0.6–1.6)0.99
 Other1.0 (reference) 1.0 (reference) 
Education    
 Less than HS diploma1.2 (0.8–1.7)0.401.3 (0.8–2.2)0.22
 HS or some college1.1 (0.8–1.4)0.631.0 (0.7–1.4)0.97
 College degree or more1.0 (reference) 1.0 (reference) 
ASA score    
 III or IV0.9 (0.7–1.2)0.440.9 (0.7–1.2)0.50
 I or II1.0 (reference) 1.0 (reference) 
Deyo-Charlson Index    
 00.9 (0.7–1.2)0.410.8 (0.6–1.1)0.24
 11.1 (0.8–1.4)0.770.8 (0.6–1.2)0.39
 >11.0 (reference) 1.0 (reference) 
Presurgical 5-year max weight loss, per 1 kg1.09 (1.06–1.12)< 0.0011.1 (1.07–1.13)< 0.001
Presurgical 5-year max weight gain, per 1 kg§1.0 (0.98–1.02)0.840.98 (0.96–1.01)0.19

DISCUSSION

The multivariable-adjusted odds of gaining ≥5% of baseline body weight over the 5-year postoperative period for persons with TKA were 60% higher than for peers who did not undergo TKA surgery. The odds of clinically important weight gain more than doubled, relative to that of non-TKA peers, for persons who underwent at least 1 additional arthroplasty procedure during the 5-year postoperative period. For TKA patients, those who were ages <70 years and those who experienced greater weight loss prior to surgery were at greatest risk for clinically important weight gain after TKA.

The finding of an increased risk of postsurgical weight gain for younger patients takes on added importance because TKA surgery for younger patients is becoming more common in the US (38) and internationally (39, 40). Kurtz and colleagues estimate that by 2016, the majority of persons undergoing TKA will be ages <65 years (38). Heisel and colleagues also found that younger patients gained more weight following TKA than older patients (41). Epidemiologic data also suggest that persons in their 50s and 60s gain more weight than persons in their 70s and older, and our data are consistent with this finding (12).

Comparisons of our findings to the literature need to account for time since surgery. For example, our findings were similar to those reported by Zeni and Snyder-Mackler, who reported that 2-year average postoperative weight gain among TKA patients was 2.5 kg (10). Our patients gained an average of 1.5 kg 2 years after surgery. Others reported essentially no weight change 2 years following knee arthroplasty in a sample of 188 overweight and obese patients (15).

At 1 year, our patients gained an average of only 0.23 kg. Others found similar very small increases in body weight 1 year after surgery (14, 17, 18, 42). Patients should be advised that the risk for clinically important weight gain increases, particularly after the first postsurgical year. Relative to the population, weight gain also is more likely for persons who undergo additional arthroplasty surgeries. The subsequent surgery likely reduces physical activity, increases pain and complications (43), and leads to additional weight gain. Our findings contrast with a recent report of weight loss for obese persons following hip arthroplasty (44).

Our study also appears to be the first to indicate that preoperative weight loss prior to TKA surgery is an important independent risk factor for postsurgical weight gain. Intentional weight loss is known to frequently lead to subsequent weight gain (45, 46), and we suspect this was the case in our sample. Overweight and obese patients preparing for TKA are frequently encouraged to lose weight prior to surgery, but in our experience, little instruction is typically provided for how to maintain weight loss. Evidence suggests preoperative weight loss will reduce the risk of complications and potentially revision surgery (43), and while presurgical weight loss may aid in enhancing early recovery, our data suggest at least some of the weight is gained back after surgery. Weight fluctuation is known also to increase the risk of a variety of comorbidities (47, 48). Effective weight control strategies to reduce weight in overweight and obese patients and to maintain weight loss in those with substantial preoperative weight loss should be the target of future research in this population. Our data suggest that patients ages <70 years who have lost a substantial amount of weight preoperatively and are particularly at risk for important postoperative weight gain are particularly vulnerable and should be the risk group targeted for future research.

Our study has several strengths, most notably a 5-year followup period, a large sample size, and an age- and sex-matched population-based comparison group. There were also some notable limitations. Loss to followup was fairly substantial, with more than one-third of patients and almost half of the REP subjects with missing weight data at 5 years. Although differences among those with complete and those with missing data were not extensive (Table 2), missing data combined with potential unmeasured confounding may have influenced our findings. For example, weight changes may have been related to unmeasured factors such as smoking status, medications, pain during recovery, and multijoint arthritis, although we did exclude persons with postsurgical weight loss surgery or cancer. We also had a higher loss to followup of persons with no comorbidities relative to persons with more comorbidities, particularly in the REP sample, and this may have influenced our findings. We also do not know whether the weight changes in our subjects were intentional or unintentional. Reasons for weight change may influence outcome following TKA surgery (49). Our REP control group weighed substantially less at baseline as compared to the knee arthroplasty group, which may have influenced our findings. We conducted a sensitivity analysis that adjusted for imbalances in baseline BMI differences among REP and TKA samples and our findings were not appreciably changed (see Supplementary Appendix A, available in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/acr.21880/abstract).

In conclusion, our study demonstrates an association between TKA surgery and subsequent clinically important weight gain relative to that seen in a population-based age- and sex-matched REP control sample. Risk factors for clinically important weight gain specifically for patients undergoing TKA were age <70 years and greater weight loss prior to surgery. The findings of this study have broad implications for the health care team providing care to a patient who has undergone knee replacement. Typically, the health care provider most concerned about weight gain is the primary care provider, with whom patients follow up regularly. Multidisciplinary weight loss/maintenance interventions particularly directed to those TKA patients who are younger and have lost considerable weight prior to surgery should be considered. Given the challenges of losing or maintaining weight over the long term, research efforts should target this subgroup of patients.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Riddle had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Riddle, Singh, Harmsen, Lewallen.

Acquisition of data. Riddle, Schleck.

Analysis and interpretation of data. Riddle, Singh, Harmsen, Schleck.

Acknowledgements

The authors wish to thank Ms Jennifer Krogman, Ms Sheila Marsh, and Ms Cara Miller for their assistance with the chart reviews.

Ancillary