Which Patients Are Most Likely to Benefit From Total Joint Arthroplasty?




To evaluate patient predictors of good outcome following total joint arthroplasty (TJA).


A population cohort with hip/knee arthritis (osteoarthritis [OA] or inflammatory arthritis) ages ≥55 years was recruited between 1996 and 1998 (baseline) and assessed annually for demographics, troublesome joints, health status, and overall hip/knee arthritis severity using the Western Ontario and McMaster Universities OA Index (WOMAC). Survey data were linked with administrative databases to identify primary TJAs. Good outcome was defined as an improvement in WOMAC summary score greater than or equal to the minimal important difference (MID; 0.5 SD of the mean change). Logistic regression and Akaike's information criterion were used to determine the optimal number of predictors and the best model of that size. Log Poisson regression was used to determine the relative risk (RR) for a good outcome.


Primary TJA was performed in 202 patients (mean age 71.0 years; 79.7% female; 82.7% with >1 troublesome hip/knee; 65.8% knee replacements). Mean improvement in WOMAC summary score was 10.2 points (SD 18.05; MID 9 points). Of these patients, 53.5% experienced a good outcome. Four predictors were optimal. The best 4-variable model included pre-TJA WOMAC, comorbidity, number of troublesome hips/knees, and arthritis type (C statistic 0.80). The probability of a good outcome was greater with worse (higher) pre-TJA WOMAC summary scores (adjusted RR 1.32 per 10-point increase; P < 0.0001), fewer troublesome hips/knees (adjusted RR 0.82 per joint; P = 0.002), OA (adjusted RR for rheumatoid arthritis versus OA 0.33; P = 0.009), and fewer comorbidities (adjusted RR per condition 0.88; P = 0.01).


In an OA cohort with a high prevalence of multiple troublesome joints and comorbidity, only half achieved a good TJA outcome, defined as improved pain and disability. A more comprehensive assessment of the benefits and risks of TJA is warranted.

For advanced hip or knee arthritis, total joint arthroplasty (TJA) is recommended (1–4). TJA ranks near the top among medical and surgical interventions in its capacity to improve quality of life (5–7). However, in Ontario, Canada, as elsewhere, wait times for TJA may be unacceptably long (8–10). Attempts to shorten wait times have focused predominantly on adding resources to improve the supply. Together with technical advances, reduced complications, and greater perceived efficacy, this has contributed to expanded indications for and a marked increase in the provision of TJAs across all ages. While the risk for serious adverse events following TJA is low (11–14), as the rates of TJA increase, perioperative risks, and longer term outcomes including revision, over an extended lifetime, must be considered and balanced against the relative gains in reduced symptoms, activity limitations, and participation restrictions, including the ability to work. Thus, efforts to reduce TJA wait times through enhanced TJA provision should be balanced with careful management of the demand for this procedure, including considering for which patients surgery is most appropriate.

Appropriate care is generally defined as that which is associated with net clinical benefit to the patient (15). Multiple definitions have been proposed to evaluate TJA benefit, including improvements in symptoms and quality of life (16–19) and patient satisfaction with the results (20–23). However, since joint pain is the most common reason people seek TJA (6, 24), at a minimum it seems reasonable that evaluation of the benefits of TJA incorporate improvement in pain. While researchers have begun to examine cut points for pain and disability that are associated with satisfaction with TJA results and/or achieving a patient acceptable symptom state (PASS) (20, 21, 25), there remains uncertainty regarding the “sweet spot” with respect to preoperative symptom levels that are associated with a high likelihood of a good outcome, if indeed one exists, and the influence of other patient factors on these cut points. This uncertainty leaves many patients, and possibly physicians, at a loss as to know how long and to what level pain and other symptoms should be tolerated before TJA is considered.

The goal of our study was to contribute to the dialogue regarding appropriate provision of TJA. In an established population cohort with hip and knee arthritis, we sought to determine the principal drivers of achievement of a “good” outcome, defined as a clinically important improvement in pain and disability. We assessed outcome using an overall measure of self-reported hip/knee pain and disability for two reasons. First, since patients undergoing TJA often have other troublesome hips and knees (5, 26) or low back pain (27), they may have difficulty attributing functional limitations, or improvements, to an intervention at a single joint. Second, we argued that “net clinical benefit” should take into consideration the patient's overall burden of hip/knee arthritis. Finally, since ∼10% of TJA recipients have inflammatory arthritis (14), we included both TJA recipients with osteoarthritis (OA) and those with inflammatory arthritis in our study.


Study sample.

Participants were members of a prospective, population-based cohort with at least moderately severe hip and/or knee arthritis. Details of cohort recruitment have been published previously (28–30). Briefly, participants were recruited between 1996 and 1998 (baseline) through a screening survey of 100% of the population ages 55 years and older residing in two regions of Ontario, Canada, one rural and one urban. Individuals were selected for cohort inclusion if they: 1) reported difficulty in the last 3 months with each of the following: stair climbing, rising from a chair, standing, and walking; 2) had swelling, pain, or stiffness in any joint lasting at least 6 weeks; and 3) indicated on a diagram that a hip or knee had been “troublesome.” In a validation substudy, 96% of the patients who met the screening criteria had hip or knee arthritis on examination and radiographs. Based on these criteria, a cohort of 2,411 individuals was established. Of these patients, 2,225 had OA, while the remainder had inflammatory arthritis, based on self-reported physician diagnosis, use of disease-modifying antirheumatic drugs, and examination. Baseline participation rates were 80.6% and 75.4% for the rural and urban regions, respectively.


Baseline and annual interviews were conducted to assess sociodemographic characteristics; body mass index (BMI), from self-reported height and weight; and hip/knee arthritis severity, using the Western Ontario and McMaster Universities OA Index (WOMAC) subscale and summary scores (31, 32), for which higher scores indicate greater severity. For each WOMAC item, respondents were asked to indicate the level of pain, stiffness, or difficulty performing activities due to their overall hip/knee arthritis. Participants were also asked about other musculoskeletal symptoms (number of painful, swollen, or stiff hips/knees in the past 3 months and persistent low back pain); self-rated general and mental health status (Short Form 36 [SF-36] general health and mental health subscales, respectively [33, 34], for which higher scores indicate better health); and comorbidity (for a list of conditions, participants indicated if they had been diagnosed by a physician and received treatment in the past year). Annual completion rates, adjusted for deceased and ineligible participants, were at least 80%. Individuals who were unable to complete assessments in any given year were offered participation the subsequent year. Survey data were linked, using unique anonymous patient identifiers (35–37), with provincial health administrative databases, enabling complete followup of all original cohort members regardless of whether or not they continued to complete assessments. Annual linkage with the Ontario vital statistics database and/or Registered Persons Database enabled verification of death status for those lost to followup.

Identification of primary, elective TJA recipients.

The Canadian Institute for Health Information Discharge Abstract Database, which records all hospital admissions, was searched using a previously defined methodology (5, 35–37) for the occurrence of hip or knee TJA from April 1, 1988 (to identify prebaseline TJAs) to March 31, 2011. Among individuals with a postbaseline, primary, elective TJA, we excluded those with a prebaseline, nonelective procedure (e.g., TJA for fracture or cancer) or those who died or underwent a second TJA procedure during the followup period, which was defined as 6 months.

Definition of good outcome.

In the absence of universally accepted criteria, our primary analysis defined a good TJA outcome as an overall improvement in hip/knee WOMAC score greater than or equal to the minimal important difference (MID), where the MID represented one-half of the SD of the difference between pre-TJA summary score and post-TJA summary score (38). Recognizing the challenges in defining important change (39), we performed secondary analyses to examine the proportions of patients with a good outcome using the following alternate definitions: 1) MID for WOMAC pain and function; 2) improvement in WOMAC subscale scores from presurgery to postsurgery greater than or equal to the minimum clinically important difference (MCID) proposed by Angst et al (40) (pain 13.4 points and function 11.9 points [maximum score 100 each]); 3) attainment of a PASS (41) for hip pain (WOMAC pain score ≥35) or disability (WOMAC function score ≥34.4) or knee pain (WOMAC pain score ≥32.3) or disability (WOMAC function score ≥31); and 4) a modification of the Outcome Measures in Rheumatology (OMERACT)–OA Research Society International (OARSI) responder criteria (42). For the latter, due to the absence of information on patient's global assessment of change, good outcome was defined as a relative change in both WOMAC pain and function scores of ≥20% and an absolute change in each of these scores of ≥10 points (maximum score 100 each). The Angst MCID definitions were derived from a rehabilitation intervention trial in OA and are thus considered a conservative estimate of the expected change in pain and function following TJA (40). For each of the definitions, we used the WOMAC scores obtained closest in date and prior to the index TJA as the presurgery values, and those obtained closest in date and after the end of the postoperative followup period as the postsurgery values.

Statistical analysis.

Presurgery characteristics were calculated using proportions, means, and medians as appropriate. Hip and knee replacement recipients were compared using t-tests for continuous variables and chi-square/Fisher's exact tests or Wilcoxon's rank sum tests for categorical variables, as appropriate. The mean change in overall hip/knee WOMAC subscale and summary scores from presurgery to postsurgery was calculated for the entire group and for hip and knee replacement recipients separately. Relative changes in symptoms were compared for hip versus knee replacement recipients using Wilcoxon's signed rank tests. The proportions that met our primary and secondary criteria for a good outcome were calculated with 95% confidence intervals (95% CIs) for the entire group and for hip and knee replacement recipients separately.

To identify the smallest number of variables with the best predictive validity, we used the following approach. First, stepwise variable selection was used to identify a logistic regression model of each size (i.e., each number of predictors, starting with a 1-variable model). Akaike's information criterion for each model was plotted against the number of predictors in the model, to determine the size of the best predictive model (43). Then, all-possible-subset regression was used to identify the best model of the selected size. Diagnostics, including a check of the residuals and tests for influential observations, were run (44). Finally, the relative risks (RRs) associated with each of the identified predictors were determined using log Poisson regression (45). The predicted probability of a good outcome (MID criterion) was then determined for a best and worst case scenario.

The following pre-TJA predictors were considered: sociodemographic characteristics (age at TJA, sex, level of education [high school or less versus postsecondary or higher], annual household income [≤$20,000 per year versus >$20,000 per year], region of residence [urban versus rural], living circumstances [alone versus with others], employment status [employed for pay–yes/no]); BMI (continuous variable or categorized as low/normal [<25 kg/m2], overweight [25–29 kg/m2], or obese [≥30 kg/m2]); arthritis severity (WOMAC pain, function, and summary scores, transformed to 0–100, where higher scores indicate greater severity); arthritis type (OA versus inflammatory arthritis); number of other troublesome hips/knees (0–3); presence of low back pain (yes/no); SF-36 general health and mental health subscales; and number of other comorbid conditions (0, 1, 2, or 3+).We additionally assessed for the effect of joint replaced (hip versus knee), year of surgery, hospital where the TJA was performed, and the duration of time between completion of preoperative and postoperative assessments and date of the index TJA. In sensitivity analyses, we re-ran our models after excluding those with rheumatoid arthritis (RA). Analyses were conducted using SAS version 9. The Women's College Research Institute Ethics Review Board approved the study.


Study cohort.

Among 2,411 baseline cohort respondents, 479 experienced a postbaseline TJA. Of these, 143 were excluded (88 with a nonelective or revision TJA and 55 who died or received a second TJA during followup). Of the remaining 336 TJA recipients, WOMAC data were available to estimate all outcome definitions for 202 (133 knee replacements and 69 hip replacements). Compared with those who were excluded from the analyses, those who were included were significantly younger (mean age 71 years versus 73 years, respectively; P = 0.007); no other significant differences were found (data not shown).

Pre-TJA characteristics of TJA recipients.

Characteristics of the 202 participants are shown in Table 1 and were similar for hip and knee TJA recipients. Mean age at TJA was 71 years, and most recipients were female (79.7%) and had a knee replaced (65.8%). WOMAC subscale and summary scores indicated moderate to severe hip/knee arthritis symptoms and disability, on average. Most had OA (93.1%) and at least 2 troublesome hips and/or knees pre-TJA (82.7%), including the joint that was replaced. More than one-half reported persistent low back pain (56.9%). Mean ± SD SF-36 general health and mental health subscale scores were 52.9 ± 22.7 and 71.7 ± 17.8, respectively. Only 30.2% had no other health problems, and 37.6% were obese.

Table 1. Presurgery characteristics of TJA recipients*
 Overall (n = 202)Hip replacement recipients (n = 69)Knee replacement recipients (n = 133)
  • *

    There were no significant differences between hip replacement recipients and knee replacement recipients. TJA = total joint arthroplasty; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; SF-36 = Short Form 36; IQR = interquartile range.

  • Maximum score 100.

Percent of patients10034.265.8
Demographic variables   
 Age, mean ± SD71.0 ± 7.071.4 ± 6.970.7 ± 7.1
 Sex, no. (%) female161 (79.7)54 (78.3)107 (80.5)
 Annual household income, no. (%) >$20K82 (40.6)29 (42.0)53 (39.9)
 Highest level of education, no. (%) with postsecondary education40 (19.8)18 (26.1)22 (16.5)
 Living arrangements, no. (%) living alone52 (25.7)22 (31.9)30 (22.6)
 Region of residence, no. (%) urban87 (43.1)32 (46.4)55 (41.4)
 Employed (including farming), no. (%)30 (14.9)12 (17.4)18 (13.5)
Osteoarthritis-related variables   
 Lower back pain (homunculus), no. affected/no. assessed (%)107/188 (56.9)40/65 (61.5)67/123 (54.5)
 WOMAC total score, mean ± SD46.5 ± 16.848.2 ± 16.845.6 ± 16.8
 WOMAC pain subscale, mean ± SD47.5 ± 17.649.1 ± 17.646.7 ± 17.6
 WOMAC function subscale, mean ± SD46.5 ± 17.648.4 ± 17.645.6 ± 17.6
 WOMAC stiffness subscale, mean ± SD44.0 ± 23.244.5 ± 23.243.8 ± 23.3
 Inflammatory arthritis, no. (%)14 (6.9)2 (2.9)12 (9.0)
 Number of troublesome hip/knee joints, no. (%)   
  135 (17.3)13 (18.8)22 (16.5)
  298 (48.5)24 (34.8)74 (55.6)
  330 (14.9)14 (20.3)16 (12.0)
  439 (19.3)18 (26.1)21 (15.8)
 Prebaseline primary, elective TJA32 (15.8)10 (14.5)22 (16.5)
Other health-related variables   
 SF-36 general health score, mean ± SD52.9 ± 22.754.0 ± 22.442.4 ± 22.9
 SF-36 mental health score, mean ± SD71.7 ± 17.872.7 ± 18.871.2 ± 17.3
 Number of comorbidities, no. (%)   
  061 (30.2)21 (30.4)40 (30.1)
  158 (28.7)20 (29.0)38 (28.6)
  252 (25.7)17 (24.6)35 (26.3)
  331 (15.4)11 (15.9)20 (15.0)
 Body mass index (kg/m2), no. (%)   
  Underweight/normal weight36 (17.8)15 (21.7)21 (15.8)
  Overweight90 (44.6)23 (33.3)67 (50.4)
  Obese76 (37.6)31 (44.9)45 (33.8)
Other variables   
 Days from pre-TJA survey to TJA, median (IQR)311 (133–606)266 (149–462)332 (133–657)
 Days from TJA to post-TJA survey, median (IQR)481 (318–788)571 (337–845)464 (300–743)

Changes in hip or knee pain and disability following TJA.

The mean ± SD changes from presurgery to postsurgery in hip/knee WOMAC pain, function, and summary scores were 13.9 ± 20.1, 8.9 ± 19.0, and 10.2 ± 18.05 (maximum score 100 for each), respectively (P < 0.0001 for change in pain versus change in function). The calculated MID for WOMAC summary scores was a 9.2 point improvement (95% CI 8.4–10.0). Half of the TJA recipients (53.5% [95% CI 46.6–60.4%]) met our a priori MID criterion for a good outcome (51.9% of knee replacement and 56.5% of hip replacement recipients; P = 0.55). Three individuals had pre-TJA WOMAC scores of <9; excluding these 3 individuals from our analyses did not change the results.

Predictors of good outcome following primary TJA.

Our primary analysis modeled predictors of good TJA outcome using the MID criterion. The optimal number of predictor variables was 4. The best 4-variable models all included the preoperative WOMAC summary score, number of troublesome hips/knees, and arthritis type. The 2 best models also contained a measure of overall health, with optimal fit observed for the model that included number of comorbidities (model C statistic 0.807) (Table 2). Using log Poisson regression, the probability of achieving a good outcome was greater for those with worse pre-TJA WOMAC summary scores (adjusted RR 1.32 per 10-point increase in pre-TJA score; P < 0.0001), fewer troublesome hips/knees (adjusted RR 0.82 per additional joint; P = 0.002), OA (adjusted RR RA versus OA 0.33; P = 0.009), and less comorbidity (adjusted RR per additional condition 0.88; P = 0.01) (Table 3). The predicted probability of a good outcome (MID criterion) for an individual with OA and only 1 troublesome joint (best case scenario) and for an individual with inflammatory arthritis and 4 troublesome joints (worst case scenario) are shown in Figure 1.

Table 2. Covariates selected for the best 4-variable models, based on the likelihood score statistic, in all-possible-subset analysis*
RankScore chi-square valueCovariates included in the model
  • *

    TJA = total joint arthroplasty; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; SF-36 = Short Form 36.

154.8Pre-TJA WOMAC
Number of troublesome hips/knees
Type of arthritis
Number of comorbid conditions
252.3Pre-TJA WOMAC
Number of troublesome hips/knees
Type of arthritis
SF-36 general health score
351.8Pre-TJA WOMAC
Number of troublesome hips/knees
Type of arthritis
451.8Pre-TJA WOMAC
Number of troublesome hips/knees
Type of arthritis
SF-36 mental health score
Table 3. Predictors of a good TJA outcome (MID criterion), by log Poisson regression
PredictorAdjusted RR (95% CI)P
  1. * TJA = total joint arthroplasty; MID = minimal important difference; RR = relative risk; 95% CI = 95% confidence interval; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; OA= osteoarthritis.

Pre-TJA WOMAC (per 10-point increase in score)1.32 (1.23–1.42)<0.0001
Number of comorbid conditions (per additional condition, to a maximum of 3)0.88 (0.79–0.97)0.01
Number of troublesome hips/knees (per additional joint)0.82 (0.72–0.93)0.002
Inflammatory arthritis diagnosis (reference is OA)0.33 (0.15–0.76)0.009
Figure 1.

Predicted probabilities of achieving a good outcome (minimal important difference [MID] criterion) by pre–total joint arthroplasty (pre-TJA) Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) summary score and number of comorbid conditions. a, Probabilities for individuals with osteoarthritis, 1 troublesome hip/knee (best case scenario), and 0 comorbid conditions (–·–), 1 comorbid condition (——), 2 comorbid conditions (······), or 3 or more comorbid conditions (–··–). b, Probabilities for individuals with inflammatory arthritis, 4 troublesome hips/knees (worst case scenario), and 0 comorbid conditions (–--–), 1 comorbid condition (······), 2 comorbid conditions (——), or 3 or more comorbid conditions (__ - __ -).

Predictors of good outcome using alternate definitions.

The proportions of patients in whom a good outcome was achieved using alternate definitions are shown in Table 4. For all but the PASS definitions, the number of individuals with pre-TJA WOMAC values that precluded them from fulfilling the criteria for good outcome was small (3–6 people). Greater numbers of individuals had pre-TJA scores greater than or equal to the acceptable symptom state for pain (29 knee replacement and 18 hip replacement recipients) and function (29 knee replacement and 15 hip replacement recipients). The likelihood of achieving a good outcome was significantly greater for pain than function using the Angst and PASS definitions (P = 0.0006 and P = 0.0007, respectively), but not MID criteria (P = 0.08). Good outcomes were achieved more frequently for hip than knee replacement recipients, but the difference was statistically significant only for PASS pain (P = 0.04). For all alternate outcome definitions, pre-TJA WOMAC score and number of troublesome hips and knees were significant predictors of good outcome. However, comorbidity and arthritis type were significant predictors only for outcome definitions that incorporated hip/knee function (data not shown).

Table 4. Proportion of TJA recipients with a good outcome by joint and criterion
CriterionOverall (n = 202)Hip replacement (n = 69)Knee replacement (n = 133)
  • * The minimal important difference (MID) was defined as 0.5 SD of the difference between pre–total joint arthroplasty (pre-TJA) and post-TJA Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) scores. The Angst pain and function minimal clinically important difference (MCID) criteria were defined as a change from pre-TJA to post-TJA in WOMAC pain or function subscale scores of ≥13.4 and ≥11.9, respectively. The modified Osteoarthritis Research Society International (OARSI) responder criteria defined good outcome as a relative change in WOMAC pain and function scores of ≥20% and an absolute change in both scores of ≥10. A patient acceptable symptom state (PASS) (41) for hip pain and disability was defined as a WOMAC pain score of <35 and a WOMAC function score of <34.4, respectively. An acceptable level of knee pain and disability was defined as a WOMAC pain score of <32.3 and a WOMAC function score of <31, respectively. Maximum WOMAC pain and function scores are 100 each. Values are the percent of patients with a good outcome (95% confidence interval).

  • P = 0.04 versus hip replacement. There were no other significant differences between hip replacement and knee replacement recipients.

 WOMAC summary score53.47 (46.59–60.35)56.52 (44.82–68.22)51.88 (43.39–60.37)
 WOMAC pain score53.96 (46.82–60.98)57.97 (45.48–69.76)51.88 (43.05–60.62)
 WOMAC function score48.02 (40.96–55.14)52.17 (39.80–64.35)45.86 (37.20–54.72)
Angst MCID   
 WOMAC pain score53.96 (47.09–60.83)57.97 (46.32–69.62)51.88 (43.39–60.37)
 WOMAC function score41.58 (34.78–48.38)44.93 (33.19–56.67)39.85 (31.53–48.17)
Modified OARSI40.10 (33.34–46.86)47.38 (35.56–59.12)36.84 (28.64–45.04)
 WOMAC pain score51.98 (45.09–58.87)62.32 (50.89–73.75)46.62 (38.14–55.10)
 WOMAC function score41.09 (34.31–47.87)49.28 (37.48–61.08)36.84 (28.64–45.04)

Sensitivity analysis.

When individuals with inflammatory arthritis were excluded, the optimal number of predictors was 3. The best 3-variable model included pre-TJA WOMAC score, number of troublesome hips/knees, and comorbidity (C statistic for the logistic model 0.79). The associated RRs were similar to those for the main model (data not shown).


Regardless of the definition of good outcome used, only half of the recipients of primary hip or knee replacement in our study experienced an important improvement in their overall hip/knee status following TJA. Four preoperative variables had good discriminative validity for predicting those who experienced a good outcome. Those who preoperatively had greater pain and disability, less comorbidity, OA versus inflammatory arthritis, and fewer other troublesome joints were more likely to experience a good TJA outcome.

The proportion of individuals with a good TJA outcome (53%), as we defined it, was lower than has been reported previously in surgical settings and at the level of the replaced joint (19, 21, 25, 46). Judge et al (47) used European registry data to examine symptomatic improvement 12 months after total hip replacement (THR) using a range of definitions, including those used in the present study. Using the MID definition, 70% of THR recipients were classified as responders. Escobar et al (25) found that 84% of THR recipients and 79% of knee replacement recipients achieved an acceptable symptom state postsurgery, while 95% and 86%, respectively, met the OMERACT–OARSI criteria for response.

There are several potential explanations for the higher proportion of nonresponders in this study. First, we excluded individuals who underwent a second elective TJA procedure during the followup period. Individuals who elect to have a second procedure may be systematically different from those who do not with respect to their satisfaction and outcomes with the first TJA, such that their exclusion may have resulted in an underestimation of the proportion of patients with a good TJA outcome. Second, we evaluated symptoms independent of the occurrence of an intervening TJA and did not ask participants if their symptoms had improved as a result of surgery. Thus, the potential for patient underreporting of post-TJA symptoms due to social desirability bias (48) may have been less. Finally, we evaluated a good TJA outcome a priori based on overall improvements in symptoms and disability for the hips and knees.

The high prevalence of other troublesome hips/knees, beyond the replaced joint, and comorbidity (83% and 70%, respectively) undoubtedly adversely impacted both preoperative and postoperative pain and disability. In a cohort of knee replacement recipients with a mean age of 65 years, Perruccio et al (26) similarly found that the proportion with other troublesome joints was high (57% had arthritis in the contralateral knee and 25% in one or both hips), and predicted worse patient-reported TJA outcomes. However, offsetting the above-noted effects, our criteria for a good TJA outcome were likely conservative. Escobar et al (25) recently reported mean improvements in WOMAC subscale scores that were considerably larger than those found in our study, resulting in relatively larger MID and MCID values.

While a single TJA might substantially improve pain if the replaced joint is the main contributor to the patient's pain experience, the presence of arthritis in other joints and/or comorbid conditions that impact physical function would likely attenuate the overall effect of TJA on functioning. Consistent with this hypothesis, a good outcome was achieved in a higher proportion of patients when outcome was defined solely based on improvements in pain versus function, and a good outcome was achieved in fewer patients when a good outcome was defined as an improvement in function. At a population level, one-quarter of recipients of elective TJA undergo a second elective TJA within 1–2 years of the index procedure (49). For patients with more than one symptomatic hip or knee, multiple TJA procedures, e.g., planned sequential and/or bilateral procedures, may be required in order to achieve maximal improvement in both pain and functioning. Further studies are needed to examine the influence of multiple primary procedures in such patients on achievement of good functional outcomes.

As has been shown by others (18, 19), controlling for other factors, participants with less self-reported pain and disability pre-TJA were less likely than those with greater severity to experience a good outcome. Possible explanations for this observation include the phenomenon of regression to the mean for those with high pre-TJA values and floor effects in those with low pre-TJA values. Consistent with the results of prior studies (21–23), almost one-quarter of our participants had pre-TJA values at or below PASS values for pain and disability. Potential explanations for little self-reported pain and disability include the use of analgesics, physical inactivity as a strategy to avoid pain, and the possibility that, in the setting of fluctuating symptoms, some people were reporting on a “good day.” If so, this suggests that a single, preoperative, self-reported assessment of hip/knee symptoms is inadequate.

Alternatively, our findings may reflect variability in patients' preferences for surgery and surgical decision making. In a prior qualitative study, we examined differences in perceptions of appropriateness of TJA between “baby boomers” and the elderly (24). While both groups discussed the importance of the impact of arthritis on overall quality of life as a key determinant of appropriateness for surgery, for older participants, quality of life was often tied to performance of basic functional activities, like stair climbing or walking. For younger participants, such basic functional limitations were relevant only in terms of their impact on overall enjoyment of life. Further, younger participants noted that, if TJA were desired, they would seek out a surgeon who was willing to provide it. The implication may be that over time, demand for TJA will increase at lower levels of hip/knee pain and disability. If so, there will be an even greater need for strong evidence to support the level of symptom severity at which TJA benefits outweigh surgical risks, both in the short term, as well as in the longer term, including risk of revision.

Weight was not a significant predictor of a successful TJA outcome, on its own or when added to the model that included WOMAC score, number of affected joints, number of comorbidities, and RA status. There are a number of potential explanations for this finding. First, it is possible that obesity was not significant due to patient selection for TJA. That is, surgeons may select overweight patients for surgery only if they are otherwise healthier than nonobese patients. Second, we may have missed an effect of weight if the effect occurs later than 6 months after surgery. Third, weight may truly not be related to outcome. Finally, the impact of weight might be mediated by the other factors examined. However, neither weight nor BMI was found to be related to the number of affected joints, number of comorbidities, or RA status, while for presurgery WOMAC summary scores, only a weak positive correlation with BMI was noted (Spearman's correlation coefficient 0.18; P = 0.01).

We did not evaluate patient satisfaction as a measure of TJA success. While improvements in pain and disability would be expected to correlate positively with satisfaction with surgical outcomes, Quintana et al (20) have recently shown that patients with worse preoperative symptoms and functioning (WOMAC scores) require a larger symptom improvement following TJA to be “very satisfied” with their outcome and that these same individuals achieve an acceptable symptom state at a higher or worse level of postoperative pain and disability than those who underwent surgery at lower symptom levels. Given their findings, our use of the same definition for good outcome, irrespective of pre-TJA severity, may have overestimated the proportion of patients with a good outcome among those with high preoperative scores, and underestimated good outcome in those with low pre-TJA scores.

Consistent with a previous population-based study (14), 7% of the TJA recipients in the present study had inflammatory arthritis. Controlling for other factors, individuals with a diagnosis of inflammatory arthritis were less likely to experience a good TJA outcome. However, the confidence interval around the relative risk estimate was large, reflecting the small number of patients with inflammatory arthritis. Larger studies are needed to confirm our findings, and, if confirmed, to elucidate potential modifiable determinants.

Previous studies of TJA outcomes have largely focused on explaining the variability in outcome. In contrast, our aim was to identify the smallest number of predictors with the best discriminative validity for a good outcome. From our final model, we produced simple graphs that depicted the likelihood of good outcome given the 4 key patient factors. It is premature to consider these graphs for use in clinical practice, since confirmation of our findings in much larger data sets are required, in addition to clarifying key determinants of other important TJA outcomes (e.g., return to work, health care use, surgical complications, and costs of revision surgery). If confirmed, however, such tools may be useful to assist patients and clinicians in making decisions about TJA.

The strengths of the present study include its focus on the impact of TJA on overall hip/knee symptoms and disability, and thus the impact of TJA in the context of overall arthritis burden, its population base, and consideration of a comprehensive list of predictors. However, there were also some limitations. First, our sample of TJA recipients was small, potentially limiting statistical power. Despite this, our final model had good discriminative validity and model fit. Second, our TJA recipients were older, on average, than TJA recipients overall. Thus, the generalizability of our findings to younger individuals must be confirmed. Third, we did not examine the influence of preoperative and postoperative process or system factors, e.g., the rehabilitation regimen provided, on outcome. Fourth, our analyses were at a group level, and thus caution is needed with respect to application of our findings to individual patients. Fifth, since most of our participants underwent their initial TJA between their baseline and first followup assessments, we were unable to examine whether the trajectory of pre-TJA changes in symptoms or disability may have influenced TJA outcomes. Finally, while many definitions of good outcome were considered, none has been validated for use in the context in which we used them. Future work is needed to examine the stability of estimates using other methods.

In conclusion, in a population cohort with hip/knee arthritis, approximately half of the patients experienced a good outcome as defined by improvements in overall hip/knee pain or disability following a single, primary TJA. The probability of a good outcome, regardless of the definition, was higher with greater preoperative pain and disability, less comorbidity, fewer troublesome hips or knees, and in individuals with OA versus those with inflammatory arthritis. As demand for TJA increases, there is an urgent need to identify patient factors that are associated with a high likelihood of experiencing a net clinical benefit, in particular the degree of arthritis severity at which the benefits of TJA are likely to outweigh potential risks. Better understanding of the determinants of a “good” TJA outcome will enable the development of tools to assist policy makers, patients, and clinicians in making recommendations and decisions regarding the provision of TJA.


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. Hawker 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. Hawker, Croxford, Davis, Dunn, Jaglal, Kreder, Sale.

Acquisition of data. Hawker, Jaglal.

Analysis and interpretation of data. Hawker, Badley, Borkhoff, Croxford, Davis, Dunn, Gignac, Jaglal, Sale.


We would like to express our thanks to our research staff and to the participants in our longitudinal cohort study for their efforts.