Study reference

(first author, year of publication)

Was there a representative and well-defined sample of patients at a similar point in the course of the disease?

(yes/no/unclear)

Was follow-up sufficiently long and complete?

(yes/no/unclear)

Was the outcome of interest defined and adequately measured?

(yes/no/unclear)

Was the prognostic factor of interest defined and adequately measured?

(yes/no/unclear)

Was loss to follow-up / incomplete outcome data  described and acceptable?

(yes/no/unclear)

Was there statistical adjustment for all  important prognostic factors?

 (yes/no/unclear)

Level of evidence

Polkowski 2013

Unclear*1

Unclear

Yes

Yes

Yes

No*1

B

Bade 2012

Unclear*2

Yes

Yes

Yes

Yes

Yes

B

Haverkamp 2012

Yes

Yes

Yes

Yes

Unclear*3

No*3

B

Hooper 2012

Yes

Yes

Yes

Yes

Yes

No*4

B

Jamsen 2012

Yes

Yes

Yes

Unclear*5

Unclear

Yes

B

Judge 2012

Yes

Yes

Yes

Yes

Unclear*6

Unclear*6

A2/B

Judge 2011

Yes

Yes

Yes

Yes

Unclear*7

No*7

B

Chang 2010

Unclear*8

Yes

Yes

Yes

Yes

Unclear*8

B

Riddle 2010

Yes*9

Yes

Yes

Yes

Yes

Yes

A2/B

Scott 2010

Unclear*10

Yes

Yes

Yes

Yes

No*10

B

Zeni 2010

Yes

Yes

Unclear*11

Yes

Yes

Unclear*11

A2/B

Cushnaghan 2009

Yes*12

Yes

Yes

Yes

No*12

No*12

B

Dauty 2009

Unclear*13

Yes

Yes

Yes

Yes

Unclear*13

B

Kim 2009

Unclear*14

Yes

Yes

Yes

Yes

Unclear*14

B

Nilsdotter 2009

Yes

Yes

Yes

Yes

Yes

Unclear*15

B

Forsythe 2008

Yes

Yes

Yes

Yes

Yes

No*16

B

Franklin 2008

Unclear*17

Yes

Yes

Yes

No*17

Yes

B

Lundblad 2008

Yes*18

Yes

Yes

Yes

Yes

Unclear*18

B

Singh 2008

No*19

Yes

Unclear*19

Yes

No*19

No*19

B

Lingard 2007

Yes

Yes

Yes

Yes

Yes

Yes

A2

Solomon 2006

Unclear*20

Unclear*20

Yes

Unclear*20

No*20

Yes

B

Study reference

Study characteristics

Patient characteristics

Prognostic factor(s)

Follow-up

Outcome measures and effect size

Comments

Polkowski 2013

Case-cohort study

University hospital referral centre; St Louis, USA

Some authors receive commercial benefits [but no obvious conflicts of interest]

Inclusion criteria: patients referred to the institution for painful knee arthropl during a 1-year period (second opinion)

Total=132 symptomatic knees

Exclusion: previous revision, unicomp, patellofem arthrop (n=6); treated with revision surgery (n=36); specific diagnosis based on objective physical examination or radiographic finding but elected nonoperative treatment (n=41)

Rest=group A

49 knees (39%) with high degree of pain and dissatisfaction in absence any objective clinical, radiographic, or serologic abnormality; all have moderate to severe pain, and were therefore dissatisfied with the result of their TKA

A/B/C/D (see column to the right)

Sex (male %)

29/42/35/35

Mean age

66/64/66/65

Mean BMI

32/33/32/31

Prior arthroscopy (%)

40/26/18/9

A vs D stat sign diff;

A vs C and C vs D near stat sign diff

Grade of osteoarthritis (OA)

Measurement: preoperative weightbearing AP radiographs; Kellgren and Lawrence (K-L) grade determined by blinded reviewer

Severe OA = K-L grade-4

Early grade OA = K-L grade 1 or 2

Analysis: incidence of pre-operative early grade OA in group A vs B/C/D

Comparison groups:

From local registry

Group B

 consecutive 100 patients undergoing prim TKA by senior surgeon for OA in same calendar year

Group C

patients with asymptomatic TKAs defined as clinical pain rated none on KSS or 0 0n VAS returning for 1-5 year followup after prim TKA for OA during the same time period

Group D

 symptomatic TKAs with KSS pain scores (moderate or worse; 4 on VAS) returning for 1- to 5-year followup after primary TKA for OA during the same time period.

Groups are not comparable with respect to % prior arthroscopy [suggesting differences in indication for TKA; see Comments]

Endpoint of follow-up: not clearly defined [probably 1-4 years post-op)

For how many participants were no complete outcome data available?

weightbearing AP radiographs:

11/49 knees (22%) group-A

0% in B/C/D

Reasons for incomplete outcome data described? No

Association of low-grade OA (pre-op) with painful TKA with no other identifiable abnormality

Comparison of cases (group-A) to ‘controls’ (group B/C/D)

Incidence of low-grade OA

K-L grade 1 or 2 (% total)

A = painful TKA with no

abnormality defined

50%

B = consecutive primary TKAs

during the same year

5%

C = normal TKAs with no pain

at followup

6.3%

D = normal TKAs with some

pain at followup

10%

A vs B/C/D: p< 0.001

Authors conclude that high %  patients referred for unexplained pain after TKA had early-grade OA preoperatively.

‘Patients undergoing TKA for less

than Grade 3 or 4 OA should be informed that they may be at higher risk for persistent pain and

dissatisfaction’

Note: total number of patients with low-grade OA that undergo TKA is unknown (group-A = ‘exposed’cases only); but ‘given 19 such patients

identified in a single surgeon’s practice during 1 year, 13% of symptomatic total knees referred

for evaluation (19 of 142), it seems likely that this situation not rare’

Note: group C (painless TKAs) and group D (painful registry TKAs) do not show a clear difference in the incidence of early-grade OA >> this may be due to a difference in the indications for TKA used in patients from group A vs C and D. In C/D more strict radiographic criteria may have been used resulting in few low-grade OA patients being selected for TKA

‘Performing a TKA on early-grade arthritis is not only counter to most generally accepted indications, but based on these results, there is strong circumstantial evidence that these patients are at risk for pain and dissatisfaction after TKA, even when performed technically well’

Bade 2012

Retrospective cohort; selected patients from trial NCT00224913 (for whom pre-operative data were available)

University hospital; USA

Non-commercial funding; conflicts of interest not stated

Inclusion criteria: Kellgren-Lawrence grade 3-4 OA; scheduled for TKA

Exclusion criteria: max pain in non-operative knee > 4 out of 10; symptomatic arthritis in any other lower extremity joint; CVD impairments, neurol impairments, uncontrolled hypertension, uncontrolled diabetes, BMI>40

All subjects underwent unilateral TKA, posterior cruciate ligament sacrificing

condylar implants, patellar resurfacing

Followed by

 inpatient post-operative care, homecare physical therapy, outpatient physical therapy program

45.4% male

64.8 (SD 9.2) years

BMI 30.4 (4.4)

6MW 467 (119)

SCT 19.9 (9.0)

TUG 10.0 (2.7)

3 functional performance-based outcome measures: Timed Up and Go (TUG), Six-Minute Walk (6MW) and Stair Climbing Test (SCT)

Prognostic factors (2 weeks pre-op) deduced from:

self-reported questionnaires (KOS-ADLS; SF-36 [-PCS, -MCS]; Global rating [how would you rate your overall functional ability from 0-100?]);

clinical/anthropometric measures (flexion, ROM, quadriceps strength, age, BMI, gender)

Analysis: CART (classification and regression trees) with TUG, 6MW, or SCT as outcome

Endpoint of follow-up: 6 months post-operatively

For how many participants were no complete outcome data available?

Not stated [probably nil]

Reasons for incomplete outcome data described?

Not stated [probably not relevant]

Timed Up and Go (TUG), Six-Minute Walk (6MW) and Stair Climbing Test (SCT) at follow-up (6 months)

TUG (mean; SD)

7.9 ± 1.8 sec

Best predictor: TUG at baseline

preop TUG ≥ 10.1 sec and aged ≥ 72 years had poorest performance: 10.8 [10.0;11.6]

preop TUG > 7.6 secs best outcome: 5.8 secs [5.4;6.2]

6MW

543 ± 118 m

Best predictor : 6MW at baseline

preop 6MW <314 m poorest outcome: 363 m [297;428]

preop 6MW ≥ 668 m best outcome: 785 m [723;848]

SCT

12.9 ± 5.0 sec

Best predictor: SCT at baseline

preop SCT ≥ 17 sec and who scored < 40 on SF-36 MCS poorest: 22.4 sec [12.7;32.1]

preop SCT > 17 sec best outcome: 10.3 sec [9.5;11.1]

Authors conclude that Poorer performance preoperatively on the 6MW, SCT, and TUG, was related to poorer performance in the same measure after TKA. Age and decreased mental health were secondary predictors of poorer performance at 6 months on the TUG and SCT, respectively.

‘These measures may help further develop models predicting thresholds for poor outcomes after

TKA.’

Note: some patients may tolerate their pain and delay joint replacement, but in so doing, their functional status deteriorates. Because functional performance prior to surgery is predictive of functional performance after surgery,, delaying surgery may

predispose them to a poorer functional outcome following TKA.

Note: SF-36 PCS and KOS-ADLS were not significant predictors at 6 months

on SCT, 6MW, and TUG.

Note: limitations of study > small sample size, strict inclusion criteria. Note that the ordered categories in the CART analysis do not imply clinically relevance. Cut-off values will change with larger sample size

Haverkamp 2012

Prospective cohort

Orthop dept; The Netherlands

Funding not stated; conflicts of interest not stated

Inclusion criteria: not stated; consecutive patients Oct 2003 – Nov 2005; prim TKA or THA

Exclusion criteria: not stated

189 TKAs

189 THAs

306 patients could be tracked

155 TKA patients analysed

Age, overall (TKA+THA)

68.3 (SD 8.7)

For TKA:

Nocturnal pain

None/mild 45.0%

Moderate 32.9%

Severe/extr 23.5%

Pain at rest

None/mild  45.6%

Moderate 37.6%

Severe/extr 18.1%

Pre-op: nocturnal pain, pain at rest (5-item Likert scale); VAS pain during activity, VAS pain during rest, WOMAC

Post-op: WOMAC, VAS pain at rest, VAS pain during activity, VAS satisfaction

Analysis: ANOVA (no adjustment for confounders) with WOMAC or VAS satisfaction as outcome

Endpoint of follow-up:

average 2.3 years (SD 0.4)

For how many participants were no complete outcome data available?

39/189=20% lost to follow-up

Reasons for incomplete outcome data described?

Not stated

VAS satisfaction, VAS pain, WOMAC; one-way ANOVA (no adjustment for confounders)

VAS satisfaction

No stat sign influence of pre-op nocturnal pain or pain at rest

VAS pain

Stat sign influence of pre-op nocturnal pain or pain at rest

WOMAC

Stat sign influence of pre-op nocturnal pain or pain at rest

More nocturnal pain and pain at rest preoperatively resulted in a less favourable outcome at follow up although more improvement was achieved

[no statistics provided; change in VAS pain or WOMAC not shown]

Authors conclude that the presence of pain at night and pain at rest in symptomatic osteoarthritic patients

results in worse VAS-pain and WOMAC scores, but with similar amounts of satisfaction at follow up.

Note: statistics lacking; simple analysis i.e. no adjustment for confounders

Hooper 2012

Prospective analysis; New Zealand joint registry

Nationwide registry; New Zealand

Non-commercial funding; conflicts of interest not stated

Inclusion criteria: TKA or THA patients in New Zealand; with ASA physical status classification (recorded since 2005)

Exclusion criteria: none [all TKAs and THAs are registered; 97% capture rate]

Year 2005-2008

18.434 TKA patients

ASA class (%) / Age

1 11% / 62.3

2 63% / 67.8

3 25% / 70.7

4 0.5% / 72.7

(5 inappropriate for elective joint surgery)

Pre-op: ASA; American Society of Anesthesiologists (ASA) physical

status classification

1= normal healthy patien

2= mild systemic disease

3= severe systemic disease

4= severe / constant threat to life

Recorded pre-operatively in the registry

Post-op: mortality within 6 months; revision rates (time to first revision during 4-5 years followup); Oxford Knee Score (OKS) at 6 months

Recorded postoperatively in the registry

Analysis: ANOVA (mortality, OKS); survival analysis adjusted for age/sex/hospital [public or private] (time to first revision)

Endpoint of follow-up: 6 months; survival analysis over 4-5 years (time to first revision)

For how many participants were no complete outcome data available?

Not stated

Reasons for incomplete outcome data described?

Not stated

Mortality within 6 months; revision rates (time to first revision during 4-5 years followup); Oxford Knee Score (OKS) at 6 months

Mortality (%; [95% CI])

per ASA class / average class age

1:  0.12 [0.01;0.43] / 62.3 y

2:  0.29 [0.20;0.42] / 67.8 y

3:  1.01 [0.73;1.37] / 70.7 y

4:  2.44 [0.30;8.53] / 72.7 y

All class differences statistically significant (p<0.05) except 1 vs 2, and 3 vs 4

Revision

Rates range from 1.2% to 1.6% at two years follow-up

No statistically significant differences between ASA classes 1, 2, and 3 (number of patients in class-4 too low to allow analysis); [corrected for age and  sex]

OKS (mean; SD)

per ASA class

1:  38.5 (0.4)

2:  37.5 (0.2)

3:  35.9 (0.3)

4:  34.6 (2.0)

Diff 1 vs 3 = 2.7 [1.7;3.7]

Diff 2 vs 3 = 1.7 [1.1;2.3]

Class differences statistically significant (p<0.001) for 1 vs 3, and 2 vs 3 [after correction for age, sex, type of institution)

Authors conclude that the ASA physical status score can be used as a predictor of postoperative mortality and functional status following knee arthroplasty

Note: 6 month mortality risk used as a measure of post-operative mortality risk (because the in-hospital mortality rate for TKA is extremely low)

Note: class difference in OKS score is clinically relevant at a 3 points difference

Note: limitations > only limited adjustment for potential confounders; pre-operative OKS not recorded [risk of confounding]. Mortality rates were not corrected for age?

Jämsen 2012

Prospective analysis; hospital registry

Tertiary care center (hospital for joint replacement); Tampere, Finland

No external funding; declared no conflicts of interest

Inclusion criteria: publicly funded primary TKA and THA in hospital district between Sept 2002 – Jan 2008

Exclusion criteria: prior open surgery of the involved joint; patients with TKA and THA in the same operation session

8083 TKA+THA

6372 patients

Antibiotic prophylaxis in all but  three patients

Median age (range)

72.2 (38.3-97.1)

Male sex

27.8%

BMI

<25 12.7%

25-29 39.6%

30-34 31.1%

35-39 11.7%

>40 4.9%

Prosthesis type

cemented 93.8%

hybrid 6.2%

uncemented 0.1%

Pre-op: demographics, preoperative health, operative details collected prospectively in joint replacement database; diagnosis diabetes based on whether patient was authorized to receive reimbursement for diabetes medications, or had purchased such medications, before the surgery

Post-op: periprosthetic joint infection during first postoperative year (hospital infection register)

Analysis: multivariate binary logistic regression; adjusted for age, sex, ASA risk score, arthroplasty site (hip or knee), BMI, and diabetic status

preliminary analyses indicated that effects of obesity and diabetes on infection rate were comparable for TKA and THA; data were therefore combined (subgroup analyses were also performed)

Endpoint of follow-up: 1 year

For how many participants were no complete outcome data available?

Not stated

Reasons for incomplete outcome data described?

Not stated

Adjusted odds ratio (OR; 95% CI) for periprosthetic joint infections in first post-operative year

adjusted for age, sex, ASA risk score, arthroplasty site (hip or knee), BMI, and diabetic status

52 periprosthetic joint infections

THA 0.64%

TKA 0.79%

TKA and THA analysed together:

multivariate model

BMI

<25    1

25-29 1.01 [0.32;3.21]

30-34 1.76 [0.56;5.56]

35-39 0.83 [0.17;4.01]

>40    6.41 [1.67;24.59] morbid obese

Stat sign OR for morbid obesity

TKA: 6/156 in morbid obese (3.85% [1.77; 8.14] ; adjusted OR= 8.06 [1.61; 40.37]

In obese patients, infection associated (stat sign) with male sex, a higher ASA risk score and with diabetes

Diabetes diagnosis

12.3% diagn diabetes at surgery

None                   1

Before surgery  2.31 [1.12;4.72]

After surgery     1.23 [0.41;3.72]

Stat sign OR for pre-op diagn diabet

TKA: 1.59% [0.84; 2.99] in preop diagn diabetes compared with 0.66%

[0.43; 0.99%] in patients without preop diagn diabetes (adjusted OR= 1.85 [ 0.75;4.58]) [not statist sign]

In patients with a preop diagn diabetes, infection associated (stat sign) with male sex  and with higher BMI

Preop glucose

<6.1      1

6.1-6.8  0.96 [0.19;4.87]

>6.9      2.25 [0.60;8.50]

Trend for increased infection rate in patients with increased blood glucose

Authors conclude that diabetes and morbid obesity increased the risk of periprosthetic joint infection following primary hip and

knee replacement.

‘The benefits of joint replacement should be carefully weighed against the incidence of postoperative

infection, especially in morbidly obese patients.’

Note: periprosthetic joint infection ratewas close to 10% in morbidly obese patients with diabetes,

which raises the question of whether performing joint replacements

in patients who are both diabetic and morbidly obese is appropriate.

Judge 2012

Prospective cohort

EOC Elective Orthop Centre [covering 1.5 million people in West-London]; London, UK

Non-commercial funding; declared no conflict of interest

Inclusion criteria: prim unilateral (unicomp or total) knee replacement

Exclusion criteria: revision; second prim replacement on the other side; bilateral operations

3608 prim TKR

Years 2005-2008

1991 (55%) completed pre-op and 6 month post-op OKS

N=1991

Mean age (SD)

71.3 (9.4)

Sex (%Male)

37.9

Mean BMI

29.4 (5.3)

IMD 2004 deprivation

medium (IQR) 11.0 (6.8–-18.8)

Main diagnosis (%)

OA 94.4

RA 2.8

Other 2.8

Operation type (%)

TKR 92.0

Unicomp 8.0

Baseline OKS

40.1 (8.1)

EQ5D anxiety (%)

Not  52.7

Moderate 41.8

Extremely 5.5

Pre-op: age, gender,height and weight, side of surgery, primary diagnosis, type of operation and ASA; Index Multiple Deprivation (IMD) 2004: measure of social deprivation. Patient- completed OKS and QoL (EQ5D) questionnaire

Post-op: Patient- completed OKS and QoL (EQ5D) questionnaire

(1)     6 months OKS (total, pain, function) – ANCOVA (Beta values)

(2)     Clinically important OKS [defined in relation to satisfaction to surgery] – logistic regression (OR)

Multivariate models; with  age, sex, BMI, IMD 2004, side of surgery, primary diagnosis (OA, RA, other), operation type (TKR,

unicompartmental), ASA grade (1, 2, 3 and 4), pre-operative OKS, pre-operative EQ5D anxiety/depression

question and year of surgery

Multiple imputation of missing values

Endpoint of follow-up: 6 months

For how many participants were no complete outcome data available?

44.8% (89.8% of these responded to the satisfaction question required for analysis (2))

Reasons for incomplete outcome data described?

Yes (responders were older, had lower BMI, were less likely to suffer from anxiety/depression; no difference in baseline pain and function)

6-months OKS (total; pain; function): regression coefficients (analysis (1)), Odds ratio (OR; analysis (2)); 95%CI

Analysis (2): clinically significant outcome (predictors of PASS score)

OR-values (95% CI)

1^st row=total OKS

2^nd row=OKS-pain

 3^rd row=OKS-function

OR>1.0 = better outcome

Baseline OKS (5-unit increase)

1.52 (1.40, 1.66) total

1.81 (1.52, 2.17) pain

2.08 (1.82, 2.39) function

Age (5-unit increase)

0.94 (0.88, 1.00) total

0.98 (0.92, 1.05) pain

0.93 (0.87, 0.99) function

Sex (Female)

0.92 (0.72, 1.17) total

1.02 (0.81, 1.28) pain

0.79 (0.62, 1.00) function

BMI

0.90 (0.80, 1.01) total

0.94 (0.84, 1.06) pain

0.92 (0.82, 1.04) function

Log IMD 2004

0.73 (0.62, 0.87) total

0.80 (0.68, 0.94) pain

0.76 (0.64, 0.89) function

Diagnosis-Primary OA

1.00 total

1.00 pain

1.00 function

Diagnosis-RA

2.17 (1.02, 4.60) total

2.33 (1.03, 5.29) pain

1.56 (0.73, 3.31) function

EQ5D-Not anxious

1.00 total

1.00 pain

1.00 function

Moderately

0.80 (0.63, 1.01) total

0.67 (0.54, 0.84) pain

0.77 (0.61, 0.97) function

Extremely

0.70 (0.42, 1.18) total

0.51 (0.31, 0.84) pain

0.77 (0.45, 1.29) function

strongest (stat sign) predictors were baseline OKS, area deprivation, diagnosis of RA, and anxiety-depression

The authors conclude that the strongest determinants of outcome include pre-operative pain/function—those with less

severe pre-operative disease obtain the best outcomes; diagnosis in relation to pain outcome—patients

with RA did better than those with OA; deprivation—those living in poorer areas had worse outcomes; and anxiety/depression—worse pre-operative anxiety/depression led to worse pain.

Differences were observed between predictors of pain and functional outcomes. Diagnosis of RA and anxiety/depression were associated with pain, whereas age and gender were specifically associated with function.

People living in poorer areas are less likely to consult a GP or seek help

at a later stage of disease, and have

worse outcomes than if they had sought help at an earlier stage

BMI was not a clinically important predictor of outcome. ‘We can conclude that in relation to patient-

reported outcomes of TKR, age and BMI should not be a barrier to surgery.’

Note: predictors explain <20% of the variability in the patient-reported outcomes of TKR; other factors (not included in the analysis) may explain a greater proportion of the variability, such as comorbidities and surgical technique.

Judge 2011

Prospective cohort

EOC Elective Orthop Centre [covering 1.5 million people in West-London]; London, UK

Non-commercial funding; declared no conflict of interest

Inclusion and exclusion criteria: see Judge 2012

1784 TKR (1523 THR)

Patients routinely complete OHS or OKS questionnaire pre-op and 6 months post-op

Overall satisfaction at 6 months (VAS): 0 = not satisfied and 100 = very satisfied

Analysis: Spearman’s  rank  correlation  coefficients; association  between pre-op OKS and satisfaction ;stratified by baseline OKS; satisfaction scoresimplified into binary variable using cut-off of >50 (identified 89% as being satisfied after TKR)

Endpoint of follow-up: 6 months

For how many participants were no complete outcome data available?

50%

Reasons for incomplete outcome data described?

Yes (responders were slightly older)

Association pre-op OKS with 6-month post-op patient satisfaction

Spearman’s rank correlation

0.04 [-0.01;0.08] not stat sign

Kruskal-Wallis test

No evidence of a difference in median satisfaction scores by categories of baseline OKS

 Chi-squared test

using categories of baseline OKS and binary satisfaction

Pre-op OKS / satisfaction (%)

<20 / 87.4%

>20 / 91.2%

p=0.01

Pre-op OKS / satisfaction (%)

low (0-16)  / 86.7%

medium (17-23)  / 90.0%

high (24-47)  / 91.1%

p=0.04

ROC analysis

AUC= 0.56 [0.51;0.60]

pre-operative OKS does not accurately predict which patients will be satisfied  or  dissatisfied 6 months after surgery (although the worst pre-operative scores were more dissatisfied)

Authors conclude that pre-operative Oxford scores have no predictive accuracy in distinguishing satisfied from dissatisfied patients.

‘This evidence does not support their current use in prioritising access to care.’

‘Currently  those with  a  higher OKS would  not  be prioritised for surgery according to SWLECI criteria, even

though they would have better satisfaction outcomes’

‘The data suggest  that  this  would  unfairly  discriminate  against patients with higher pre-operative  scores, who would do well  post-operatively  with  significant  relief  of  pain  and improved mobility. In fact, our data after TKR support surgery  in  those with higher  scores so as  to avoid poor outcomes in patients with very low pre-operative scores.’

Chang 2010

Prospective (?) cohort

University hospital; Seoul, South Korea

No external funding; authors declare no commercial interests

Inclusion criteria: consecutive patients; primary TKA; osteoarthritis

November 2003 to March 2005

Exclusion criteria: other diagnosis such as rheumatoid arthritis, periprosthetic infection

Exclusion during followup (55 TKAs): death or significant medical problems unrelated to the surgery, spine / hip fractures); not visiting for 1-year follow-up (12 TKAs)

438 TKAs (276 patients)

55 excluded during fup

Study population:

383 TKAs

240 patients

Sex

230 female (96%)

10 male (4%)

mean age

68.8 (range 54-82)

mean BMI

26.2 (range 18–37)

Laterality

Bilateral 75%

Unilateral 25%

Radiographic OA score

5.4 points (SD 1.7)

Preop WOMAC (SD)

Pain 11.8 (4.2)

Stiffness 4.9 (2.0)

Function 42.2 (12.7)

Total 58.9 (17.3)

Clinical and demographic data collected by independent investigator:

Preop: WOMAC scales; radiographic evaluation (modified Ahlback radiographic scoring system)

Post-op: WOMAC and patient satisfaction (4-point system of British Orthopaedic Association: enthusiastic, satisfied, noncommittal, and disappointed)

Patient satisfaction dichotomized (satisfied= enthusiastic or, satisfied versus dissatisfied= noncommittal or disappointed)

Analysis: linear regression (WOMAC); logistic regression (patient satisfaction)

'When significant associations were found between possible confounders and postoperative outcomes, confounding effects were controlled by multivariate regression analyses. These factors included gender, BMI, unilateral or bilateral, type of implant, postop decrease in ROM' [note: in the analyses presented the exact multivariate adjustment is unclear; also note that a post-operative parameter was entered into the models, post-operative decrease in ROM]

Endpoint of follow-up: 1 year

For how many participants were no complete outcome data available?

17% (patient satisfaction); 55 TKAs were excluded during follow-up

Reasons for incomplete outcome data described?

No

NS= statistically not significant

Associations between age, pre-op severity of OA, pre-op WOMAC (pain, function) with 1-year post-op WOMAC (pain, function), and with patient satisfaction; regression coefficients (p-values), odds ratio (OR; 95% CI)

post-op WOMAC pain

Regression coefficients (p-value)

Age 0.101 (NS)

OA score -0.024 (NS)

Preop  WOMAC pain 0.189 (p= 0.002)

Preop  WOMAC func  -0.031 (NS)

post-op WOMAC function

Age 0.232 (p<0.001)

OA score -0.078

Preop  WOMAC pain  0.041

Preop  WOMAC func  0.205 (p<0.001)

post-op patient satisfact

OR [95% CI]

Age 0.931 [0.849;1.021] (NS)

OA score 1.119 [0.833;1.503] (NS)

Preop  WOMAC pain

 0.866 [0.730;1.027] (NS)

Preop  WOMAC func

1.018 [0.960;1.080] (NS)

Subgroup analysis: patients meeting all 3 selection criteria versus those only meeting 1 or 2. Criteria: age > , severe radiographic OA (modified > 4.5), severe pre-op symptoms (WOMAC total score < 50%)

49.6% met all 3 criteria

5.2% did not meet age criterion

13.8% did not meet radiogr criterion

20.1% did not meet funct criterion

Post-op patient satisfaction

% dissatisfied

‘typical patients’ (meeting 3 criteria)

8.7% (14 out of 161 patients)

Not meeting age criterion

0% (0 out of 16 patients)

Not meeting radiographic criterion

8.9% (4 out of 45 patients)

Not meeting symptomatic criterion

3.0% (2 out of 66 patients)

Differences are not statistically significant

Authors conclude that older age was associated with worse postoperative functional outcomes. Worse preoperative WOMAC pain and function scores were associated with worse postoperative WOMAC pain and function scores.

The radiographic severity of OA had no significant associations with any of the postoperative outcomes (WOMAC; patient satisfaction).

Compared to typical candidates, the groups of patients who did not meet one of the selection criteria did not have worse postoperative outcomes.

The group of younger age and the group with less severe preoperative

symptoms were more likely to have better postoperative outcomes.

Note: not clear to what extend results were adjusted for potential confounders; corrections appear minimal. Note that a post-operative parameter was entered into the models:  post-operative decrease in ROM

Note: interpretation of subgroup analysis is complicated by the fact that patients not meeting the indications for TKA were not included in the study. In addition, the statistical power is low (e.g. only 20 patients did not meet the age criterion)

Note: results may not apply to other patient groups; note the high % women (96%), and over 96% was satisfied with the TKA at 1 year follow-up

Riddle 2010

Prospective cohort

University hospital; Richmond, USA

Non-commercial funding; no statement on potential conflicts of interest

Inclusion criteria: long-standing knee OA; prim knee arthroplasty; attending preop educational class

Dec 2005 -  April 2008

Exclusion criteria: revision surgery

283 patients

157 selected 'on alternate weeks' by study coordinator blinded to patient measures ('because of staffing shortages and limited funding')

129 primary TKA

28 UKT

Pre-op: patient characteristics, duration of knee pain, inflammatory arthritis, pain elsewhere, comorbidity questionnaire;

Depression questionnaire (PHQ-8)

Generalized Anxiety/Panic disorder modules (PRIME-MD)

Tampa Scale of Kinesiophobia (TSK)

Short form of Arthritis Self-efficacy Scale

Pain Catastrophizing Scale (PCS): 13-item scale (0=no catastroph; 52=severe catastroph)

Post-op: WOMAC (pain, function)

Analysis: WOMAC pain, WOMAC function; dichotomized using (a) 50% improvement from baseline, and (b) MCID for WOMAC pain [4 points] and function [15 points]

GLM (logistic regression), multivariate; adjusting for: age, gender, comorb score, RA status, race/ethnicity, preop WOMAC pain score

Endpoint of follow-up: 6 months post-op

For how many participants were no complete outcome data available?

11% (did not complete 6-month questionnaires)

Reasons for incomplete outcome data described?

No

NS = statist not sign  (p> 0.05)

Associations between psychologic disorders (  ) and 6-month post-op WOMAC (pain, function; odds ratio (OR; 95% CI)

Multivariate regression; only predictors with p<0.1

post-op WOMAC pain

Model 1: change by <50%

PCS score of ≥16

OR= 2.67 [1.2;6.1] (p=0.02)

Model 2: change ≤ 4 points [MCID]

Self-efficacy (continuous)

OR= 0.80 [0.62;1.03] (NS)

TSK (continuous)

OR= 0.92 [0.85;1.01] (NS)

PCS  ≥ 16

OR= 6.04 [1.75–20.81] (p=0.005)

post-op WOMAC function

Model 1: change by <50%

PCS score of ≥16

OR= 2.18 [0.91;5.19] (NS)

Model 2: change ≤ 4 points [MCID]

None (no psychologic variables entered the model)

Authors conclude that pain catastrophizing was the only

consistent psychologic predictor of poor WOMAC painoutcome. No psychologic predictors were associated consistently with poor WOMAC function outcome.

Patients with PCS scores of 16 or greater had an odds of poor outcome

(less than 50% improvement) that was 2.67 times that of patients with PCS scores of 15 or less.

‘An intervention focusing on pain catastrophizing seems to have potential for improving pain outcome in patients prone to catastrophizing pain.’

Note: Approximately 20% of sample had symptoms of major depression and an additional 20% had symptoms of generalized anxiety disorder. Yet most did not predict poor outcome (except pain catastrophizing).

‘We suspect that most of the potentially deleterious effects of these mental health disorders are washed out by the large surgical treatment effect.’

Note: process of recruiting was not random but pragmatic (‘ alternate week approach’; potential risk of bias)

Scott 2010

Prospective cohort

Hospital; Edinburgh, UK

No external funding; Non-commercial funding; stated ‘no financial benefits’

Inclusion criteria: consecutive TKRs

No detailed inclusion or exclusion criteria stated

 2006 -2008

1290 TKRs

1213 patients

Sex

42% Male

mean age

70.1 (range 35-92)

Pre-op: SF-12 health ques-

Tionnaire;  OKS;  questions  related  to comorbidities

Post-op: at 12 months, SF-12, OKS, and patient satisfaction

patient satisfaction: ‘How  satisfied  with operated knee?’ (‘very satisfied’, ‘satisfied’, ‘unsure’ or ‘dissatisfied’); how  well  surgery  relieved  pain  in 

affected  joint, increased  their  ability  to  perform regular activities,  heavy work/sporting activity; how it met their expectations; their overall hospital experience

Analysis: univariate and multivariate ordinal  logistic regression

Endpoint of follow-up: 1 year post-op

For how many participants were no complete outcome data available?

6% (did not complete 1-year satisfaction questionnaires)

Reasons for incomplete outcome data described?

No

NS = statist not sign  (p> 0.05)

Patient satisfaction (dissatisfaction) 1-year post-op; OR (only p-values reported)

Multivariate regression

Patient satisfaction

666 (54.7%) very satisfied

325 (26.7%)  satisfied

158  (13%)  unsure

68 (5.6%) dissatisfied

>> 81%  satisfied vs 19%  dissatisfied

‘After adjusting for other factors, depression, a low SF-12 mental component score and pain in other joints were the only pre-operative factors that remained significant predictors of dissatisfaction (p<0.001)’ [no further details provided]

‘pre-operative OKS for function was borderline (p = 0.014).’

‘Of patients known to suffer from depression pre-operatively, 42 of 127 (33%) were very  satisfied, 40  (31%)  satisfied, 28  (22%) unsure and 17 (13%) dissatisfied.

Authors conclude that  significant predictors at one year included the pre-operative SF-12 mental component score, depression and pain in other  joints, the six-month SF-12 score and poorer improvement in the pain element of the Oxford

Knee Score. [note: only pre-op SF-12 and pain in other joints are relevant and supported by [poorly documented] multivariate analysis]

Note: multivariate analysis did not include potential confounders such as BMI, previous knee surgery, under

lying pathology, presence of complications

Zeni 2010

Prospective cohort

Dept Phys Ther, university; Newark, USA

Non-commercial funding; stated ‘no commercial interests’

Inclusion criteria: consecutive prim TKRs; end-stage OA; posterior cruciate ligament-sacrificing condylar implants with patellar resurfacing

Exclusion criteria: notable pain in contralat limb; arthritis involving any other lower extremity joint; cardiovascular or neurological impairments

N=105

Age

65.8

BMI

30.9

KOS-ADLS score (%)

50.4 ± 16.7

Quadric strength (N/BMI)

Oper / non-oper side

19.28 / 23.50

Active ROM

Extension 4 ± 5

Flexion 119 ± 14

Pre-op: 11 ± 8 days prior

to surgery; functional evaluations (height, weight, active ROM, bilateral quadriceps

strength); completed Knee Outcome Survey (ADL scale of the KOS); stair-climbing ability assessed

Quadriceps Strength:measured isometrically (KinCom Dynamometer);normalized to BMI; quadriceps index= ratio of strength of involved limb to the strength of the uninvolved limb (%)

Stair-Climbing Task: twelve steps; patient ascended, and descended as quickly as possible (time recorded); subject instructed to use handrail only if needed

Note: all subjects participated in aggressive postoperative physical therapy

Post-op: handrail use (yes/no) at follow-up (3 months; 2 years post-op)

Analysis: binary logistic regression for handrail use (dichotomous); adjusted for pre-op age, BMI, quadriceps index, knee flexion ROM, KOS-ADLS score, stair-climbing task (time), and use of a handrail; forward stepwise regression model

Endpoint of follow-up: 3 months, 2 years post-op

For how many participants were no complete outcome data available?

none

Associations with handrail use at follow-up; correct prediction of handrail use (%); p-values

At 2 years follow-up

Handrail use

Pre-op: 63/105 (60%)

2 years post-op: 60/105 (57%)

77% of the patients using a handrail pre-op, also used a handrail at 2 years post-op

Regression model:

Handrail use (dichotomous)

Overall % correct prediction

Step 1

Preop handrail use

p<0.001

68% correct prediction

Step 2

Preop handrail use

p=0.001

Age

p=0.002

71% correct prediction

Step 3

Preop handrail use

p=0.001

Age

p=0.001

Quadriceps index

p=0.024

74% correct prediction

Preop BMI, knee flexion ROM, KOS-ADLS, and stair-climbing task (time) did not significantly contribute to the

prediction of handrail use after total knee arthroplasty

Authors conclude that younger, stronger patients who do not use a handrail prior to unilateral TKA can expect the best outcomes in terms of ascending and descending stairs following surgery.

‘This information may provide patients with more realistic expectations after surgery and allow them to make more appropriate discharge plans.’

Note: differences in postoperative strength of the operatively treated limb at two years suggest that quadriceps strength never fully recovers after TKA in persons who require a handrail.

Note: quadriceps weakness has been implicated in reduced functional ability 12 months after surgery and is highly related to an individual’s ability to perform ADL; falls on stairs

account for the majority of fall-related injuries.

Note: average age patients was 66 years i.e. relatively young age

Cushnaghan 2009

Case-control study

Follow-up study of 657 matched pairs of patients and controls who had participated in an earlier case–

control study of knee osteoarthritis

Hospitals; Portsmouth, Southampton, Stoke-on-Trent, UK

Non-commercial grant; declare ‘no competing interests’

Inclusion criteria: residents of 3 English health districts;

> 45 years; on

waiting list for TKA; primary knee OA

1995-1997

Exclusion criteria: inflammatory cause OA

TKA patients compared to matched controls: no previous knee surgery for OA; selected from general population; matched to the patients for age, sex and general practice

Analytic sample:

363/657 matched pairs (55%) successfully contacted at follow-up; 42 excluded e.g. because no TKA operation was performed; leaving 325 cases and 363 controls

Patients/controls

Mean age (at follow-up)

69.8 / 69.9

BMI

<25      14% / 45%

25-30  52% /  44%

>30      34% /  11%

Smoking habits; Hypertension; Diabetes; Thyroid disease; OA severity (K-L); previous injury to affected knee; Heberden nodes; number of painful joint sites in hands, shoulders and knees [see paper for details]

Pre-op: interview using structured questionnaire: sex, age, smoking habits; previous injury knees, pain in hands, shoulders and hips; comorbidity

SF-36 section related to physical function, vitality and mental health

Clinical assessment (e.g. BMI); knee radiographs scored for overall severity of OA according to Kellgren and Lawrence

Post-op: self-administered questionnaire > confirmation knee replacement surgery; SF-36 -physical function, -vitality and -mental health; review hospital records (details knee surgery)

Analysis restricted to patients who underwent TKA after the baseline interview and at least 2 years before completing the follow-up questionnaire, and to controls who were not treated by TKA during the follow-up period

Analysis: Baseline predictors of change in SF-36 physical function score over follow-up were explored by multiple linear regression, using data from the subset of subjects who provided complete information on all relevant variables. [primary analysis was not paired]

Multivariate adjustment: age, BMI, smoking habits; hypertension; diabetes; thyroid disease; OA severity (K-L); previous injury to affected knee; Heberden nodes; number of painful joint sites in hands, shoulders and knees

Endpoint of follow-up:

Patient group : mean 6.9 years

Controls: mean 6.5 years

 For how many participants were no complete outcome data available?

363/657 matched pairs (55%) successfully contacted at follow-up; 42 excluded e.g. because no TKA operation was performed; leaving 325 cases and 363 controls

Reasons for incomplete outcome data described?

No

SF-36 physical function score; Mean difference [95% CI]

SF-36 physical function

TKA patients

Baseline SF-36

phys function (per 10 units)

-4.6 [-6.3;-2.9]

Female sex

-7.5 [-14.9;0.0]

Age (years)

<70 Reference

70–75 -1.2 [-10.2;7.8]

>75 -9.1 [-16.9;-1.2]

BMI (kg/m2):

<25 Reference

25-30 -1.2 [-10.2;7.8]

>30 -4.6 [-14.5;5.3]

Smoking habits:

Never smoked Reference

Ex-smoker -5.7 [-12.4;0.9]

Current -5.1 [-17.1;6.9]

Hypertension Present

-3.4 [-9.6;2.7]

Diabetes Present

3.8 [-8.5;16.2]

Thyroid disease Present

-1.9 [-8.6;12.5]

K-L OA grade

0–2 Reference

3   7.6 [-0.2;15.3]

4   5.6 [-2.8;14.1]

Previous injury to knee

2.8 [-3.8;9.4]

Heberden nodes Present

-1.1 [-10.7;8.5]

No. of painful joint sites

in hands,shoulders and knees

0 Reference

1 -4.3 [-12.5;3.9]

2 -7.3 [-15.7;1.2]

3 -6.3 [-16.5;3.8]

4+ -10.3 [-20.9;0.3]

SF-36 physical function

Controls

Baseline SF-36

phys function (per 10 units)

-3.7 [-4.7;-2.7]

Female sex

-7.0 [-12.7;-1.2)

Age (years)

<70 Reference

70–75 -7.5 [-13.9;-1.2]

>75 -14.2 [-20.5;-7.8]

BMI (kg/m2):

<25 Reference

25-30 -3.7 [-9.2;1.8]

>30 -11.1 [-20.0;-2.1]

Smoking habits:

Never smoked Reference

Ex-smoker -4.2 [-9.9;1.5]

Current -9.6 [-17.9;-1.3)

Hypertension Present

-8.1 [-14.1;-2.1]

Diabetes Present

0.6 [-9.6;10.8]

Thyroid disease Present

-5.6 [-15.5;4.2]

Authors conclude that Improvements in physical function following TKA for osteoarthritis are sustained beyond 5 years. The benefits are apparent in patients who are obese as well as non-obese, and there seems no justification for withholding TKA from obese patients solely on the grounds of their BMI.

TKA patients: better baseline

physical function, female sex and older age associated with sign worse change in physical function over follow-up.

Controls: Better baseline physical function, female sex, older age, BMI >30, current smoking and hypertension associated with sign greater decline in physical function during follow-up.

Note: Subgroup analysis

108 patients with BMI at baseline >30: median physical function score increased from 17 to 20 points over the course of follow-up

(p=0.002), whereas among the 36 controls with baseline BMI >30, the median score decreased from 61 to 25 points (p<0.001).

82 patients who were older than 75 years at baseline, the median physical function score was unchanged over follow-up at 17 points, while in the 87 controls aged 75 years or older at baseline, median physical function score deteriorated from 83 to 43 points (p<0.001).

Note: high risk of confounding. The control group may not be representative of TKA patients not undergoing surgery, i.e. comparison may not be valid. Also note that obesity was more frequent in TKA patients than in controls (34% vs 11%). Finally, note the incomplete follow-up.

Dauty 2009

Retrospective cohort

Hospital; Nantes, France

Funding, potential conflicts of interest: not stated

Inclusion criteria: all patients admitted to rehabilitation service

following primary TKA for OA

 Jan 2004 - June 2007

Exclusion criteria: arthritis, bone tumours, fractures, osteonecrosis and TKA revisions

N=244

Patients underwent physical medicine therapy in the orthopaedic service and in the rehabilitation service

Discharge on basis of 3 criteria: independent in ADL (if not help was arranged), capable of walking well enough to reduce fall risk, joint amplitude active knee flexion >90 degrees

Mean age

71.4 (SD 8.7)

Sex

40% Male

Pre-op: medical record

Post-op: active knee flexion, walking distance, IKS function score and complications (venous thromboses, acute infections TKA, poor skin closure, acute cardiac, pulmonary or digestive decompensation)

Patients with these complications and requiring transfer to another

medical or surgical service were excluded

Analysis: multiple linear regression of duration of inpatient rehabilitation; prognostic factors studied were age, gender, cohabitational status, presence of home help, etc

Endpoint of follow-up:

Hospital discharge, mean 7.5 days (SD 1.7) after surgery

 For how many participants were no complete outcome data available?

Not stated (none?)

Duration of inpatient rehabilitation; regresson coefficient (p-value)

Multivariate linear regression

Regression coefficients

Gender

2.36 (p=0.02)

Home help

-1.89 (p=0.29) NS

Previous arthroplasty

-1.78 (p=0.11) NS

Note: r2 value of 0.02 (i.e. the model explains only 2% of the variability in the duration of inpatient rehabilitation)

Authors conclude that  the duration of inpatient rehabilitation for TKA patients in the absence of complications cannot be statistically modelled from the preoperative parameters studied here.

Nilsdotter 2009

Prospective cohort

University hospital; Lund, Sweden

Funding not stated; declared ‘no conflicts of interest’

Inclusion criteria: consecutive patients on waiting list for prim TKR

Dec 1999 - April 2001

Exclusion criteria: other operative procedures (n=13), not operated on (8) , RA (2)

125 patients

23 excluded

N=102 patients with knee OA

 Sex

38% Male

Mean age

71 years (range 51-86)

comorbid conditions (mean number)

1.26 (SD 1.16)

36.5 % has 2 or more comorbid conditions

Pre-op: KOOS; SF-36; questions regarding background data e.g. comorbidity (currently treated by doctor, or treated during last

year, for back problems, lung disease, high blood pressure, heart disease, impaired circulation in the lower extremity, neurological disease, diabetes, cancer, ulcer, kidney disease, impaired vision or eye disease)

Post-op: same questionnaires at  6 months, 1 year  and 5 years

KOOS: 42 item self-administered questionnaire, 5 dimensions (pain, other disease specific symptoms, ADL function, sport and recreation function [Sport/Rec], and knee-related quality of life).

SF-36 (Acute Swedish version): generic outcome measure, 8 domains (PF, role-physical (RP), BP, general health (GH), vitality (VT), social functioning (SF), role-emotional (RE) and MH)

Analysis: univariate analyses [not shown in evidence-table] and multivariate analyses (ANCOVA) of KOOS (5 dimensions), adjusting for age, gender, comorbidity, pre-op SF-36 (PF, BP and MH)

Endpoint of follow-up: 6 months, 1 year, 5 year

 For how many participants were no complete outcome data available?

14% (at 5-year follow-up)

Reasons for incomplete outcome data described?

No

PF= physical function

BP= bodily pain

MH= mental health

Associations with subscales KOOS; regression coefficient [95% CI]

At 1 year follow-up

KOOS-pain

Female gender       0 [-8;7]

 Older age (10 y)    5 [1;10]

 Comorb                  -3 [-13;7]

KOOS-symptoms

Female gender       2 [-5;10]

 Older age (10 y)    6 [1;10]

 Comorb                  -1 [-12;10]

KOOS-ADL

Female gender       0 [-7;7]

 Older age (10 y)    2 [-2;7]

 Comorb                  -3 [-13;8]

KOOS-Sports/Rec

Female gender                12 [-13;73]

 Older age (10 y)                15 [-2;31]

 Comorb                  12 [-23;46]

KOOS-QoL

Female gender       3 [-8;13]

 Older age (10 y)    9 [3;16]

 Comorb                  -1 [-15;14]

SF-36 PF /  SF-36 BP /  SF-36 MH

No sign associations in all 5 dimensions

At 5 years follow-up

KOOS-pain

Female gender       -2 [-12;8]

 Older age (10 y)    7 [1;13]

 Comorb                  0 [-14;14]

KOOS-symptoms

Female gender       -3 [-11;5]

 Older age (10 y)    7 [2;12]

 Comorb                  0 [-12;12]

KOOS-ADL

Female gender       -4 [-13;5]

 Older age (10 y)    4 [-2;9]

 Comorb                  6 [-7;20]

KOOS-Sports/Rec

Female gender       3 [-33;38]

 Older age (10 y)                10 [-1;28]

 Comorb                  -6 [-66;55]

KOOS-QoL

Female gender       -4 [-17;8]

 Older age (10 y)    4 [-3;12]

 Comorb                  13 [-5;30]

SF-36 PF /  SF-36 BP /  SF-36 MH

No sign associations in all 5 dimensions

Authors conclude that older age to some extent predicted more post-op pain and other symptoms, however, no predictors of post-op physical function (PF) were found, indicating the difficulty of determining pre-op

who will benefit more or less from TKA [predictors in multi-variate model explain only 5-18% of variation in post-op pain and ADL function]

Note: compared to pre-op, a significant improvement was still seen 5 years post-op. However, best result was reported at 1 year, indicating a decline from 1 to 5 years after TKR. [on dimensions of the KOOS]

Note: pre-op SF-36 (PF, BP, MH) did not predict outcome after 1 or 5 years follow-up.

Note: no control group to assess natural history; BMI was not included in the multivariate model (risk of bias); only patient-reported outcomes were assessed (no objective outcome measures); multiple testing (3 follow-up times; 5 dimensions of KOOS); power probably insufficient for some dimensions of the KOOS

Forsythe 2008

Prospective cohort

University hospital; Halifax, Canada

Funding / conflicts of interest: not stated

Inclusion criteria: consecutive patients elective TKA

Jan 1999 - Jan 2000

Exclusion criteria: none stated

70 patients; 55 (79%) willing to participate

N=55

Sex

36% Male

Mean age

range 49-85 years

70.0 (m) / 68.5 (w)

Length of stay (days)

5.9 (m) / 5.5 (w)

Nr comorbidities

2.6 (m) / 1.6 (w)

Pre-op: PCS and MPQ (in relation to the affected knee), standard medical history and physical examination, approx. 1 week before surgery

Post-op: questionnaires at 3,

12 and 24 months

Catastrophizing: Pain Catastrophizing Scale (PCS); total score (PCST), rumination (PCS-R), magnification (PCS-M) and helplessness (PCS-H)

Pain: McGill Pain Questionnaire (MPQ); patients rate adjectives (Pain Rating Index [PRI]) that best describe their current pain with

a four-point scale (0=none, 3=severe)

VAS-pain: overall intensity (0= no pain, 5= excruciating pain)

Comorbidities: from the charts; absolute number of comorbidities was used as a variable

Analysis: ROC curves to determine which preop variables predicted long-term pain outcomes; patients with persistent pain compared with patients with no pain at 24 months using Mann-Whitney U test

Endpoint of follow-up: 3, 12 and 24 months.

Primary outcome at 24 months

 For how many participants were no complete outcome data available?

8-12% (8% at 2-year followup)

Reasons for incomplete outcome data described?

No

Associations between pre-op pain / number of comorbidties and post-op pain (24 months); ROC analyses

ROC curve analysis (24 months)

pre-op PCS-T predicted pain (PRI)

AUC= 0.713 (P<0.05)

pre-op PCS-R predicted pain (PRI)

AUC= 0.696 (P<0.05)

Pre-op PCS-M and PCS-H scores did not provide significant discrimination of pain outcomes

Pre-op PRI predicted pain (PRI)

AUC= 0.720 (P<0.05)

Comorbidities predicted pain (VAS)

AUC= 0.714 (P<0.05)

Authors conclude that number of comorbidities and pre-op PCS scores predicted the presence of pain at 24 months follow-up

'This may enable the identification of knee arthroplasty patients at risk for persistent postoperative pain, thus allowing for efficient administration of preoperative interventions to improve arthroplasty outcome’

Note: patients with increased wait times for TKA and worsening preoperative pain may have a worse outcome after surgery

Note: analyses were not corrected for confounders (no multivariate analyses); small study size

Franklin 2008

Prospective; registry data

Multiple orthop dept (135 orthop surgeons; 31 states); USA

Some authors received commercial funding; conflicts of interest not stated

Inclusion criteria: national registry TKR; Zimmer NexGen knee components; patients with complete 1-year outcome data

2000 - 2005

Exclusion criteria: fracture, malignancy, infection, or failure of previous TKR; operative code for revision or bilateral TKR; emergently scheduled surgery; any known postoperative complications including infection and deep venous thrombosis.

17270 patients, of which 8050 (47%) with complete 12-months outcome data

N=8050

Sex

34% Male

Mean age

67.8

Mean BMI

32

Race

70% Caucasian

4% Afric Am

3% Hispanic

Diagnosis

95% prim knee OA

mean KSS pain

37 (SD 16)

54% mod / severe pain at rest; 92% mod / severe pain walking

Pre-op and Post-op: registry contains demographic and preoperative clinical characteristics, surgeon assessment, surgical technique, and patient-reported outcomes at 12 months after TKR, including pain (Knee Society score) and physical function (physical component score [PCS] of the SF-12).

Analysis: multivariate mixture models to predict which subgroup of patients would have low or high functional gain 12 months after surgery; patient groups modeled as 2latent classes by fitting a mixture model of two normal distributions; adjustment for patient baseline predictors of post-TKR functional gain as covariates, including gender, age at surgery, BMI, MCS and PCS, osteoarthritis diagnosis, and poor quadriceps strength prior to TKR

Endpoint of follow-up: 12 months

 For how many participants were no complete outcome data available?

53%

Reasons for incomplete outcome data described?

No

Associations between pre-op predictors and physical function (SF-12, PCS) at 1 year follow-up; ‘latent class multivariate mixture model’; regression coefficients, OR [95% CI]

Change in physical function

Regression coeff [95% CI]

Male sex 0.64 [0.05;1.23]

Age (+1 year) -0.03 [-0.07;0.003]

BMI<30                   Ref.

30–39      -1.07 [-1.69;-0.45]

40–49      -1.63 [-2.58;-0.69]

>50          -1.86 [-4.31;0.60]

MCS        0.04 [0.01;0.07]

PCS          -0.84 [-0.90;-0.78]

OA           2.61 [1.22;4.00]

Quadriceps  strength

 Poor       -1.76 [-4.73;1.20]

Higher odds of less functional gain were associated with BMI over 40, lower preoperative MCS, each 5-year increase in age, poor or fair quadriceps strength, and non-OA diagnosis

Prob for low functional responders

OR [95% CI]

Age (+1 year) 1.01 [1.00;1.02]

BMI>50   2.33 [1.09;4.99]

MCS        0.95 [0.94;0.96]

PCS          0.91 [0.89;0.92]

Quadriceps strength

Fair         1.34 [1.02;1.75]

Poor        2.28 [0.79;6.61]

Age, BMI, baseline MCS and PCS, and poor or fair quadriceps strength predicted the low function responder

Group

Patients with BMI over 50 and poor quadriceps strength had more than 2:1 odds of poor functional gain after TKR

Authors conclude that increased likelihood of poor functional gain was

associated with older age, BMI > 40,

lower MCS, and poor quadriceps strength.

Note: TKR surgery consistently relieves pain due to arthritis, but one-third of patients reported limited functional gain 1 year after surgery; analyses suggest functional gain following TKR varies by demographic and clinical variables,

including pre-TKR age at surgery, gender, BMI, emotional and physical function, and quadriceps strength.

‘Future research may examine strategies to tailor postoperative rehabilitation programs to the particular needs of these patients’

‘A better understanding of the patient attributes associated with limited improvement in post-TKR

function will allow surgeons to discuss the likely functional outcome in individual situations in order to assure patients’ expectations are satisfied’

Note: <50% of patients had complete outcome data (risk of bias; people with 12-month data may represent those with best outcomes ..); modelling approach assuming 2 classes with normal distributions, may ‘inflate’ risk ratios

Lundblad 2008

Prospective cohort

Hospital; Stockholm, Sweden

Non-commercial funding; conflicts of interest not stated

Inclusion criteria: caucasian Swedish patients; scheduled for

TKR for OA

Exclusion criteria: none stated

N=69

Sex

49% Male

Mean age

68 years (range 40-80)

Duration knee pain

8.5 years (range 1-25)

NSAIDs were withdrawn between 3-14 days prior to surgery (: ‘routine practice’)

Pre-op (day before surg): total duration knee pain and current intensity at rest (sitting, standing or lying down) and with movement (during walking); using VAS, and using Pain Matcher; pain threshold (using Pain Matcher)

Pain Matcher: exposes patient to an electrical stimulus which matches the intensity of the actual pain; provides a numerical value to a level of pain >> sensation threshold= first noticeable sensation; pain threshold = perceived signal as painful; matched pain = intensity of pain same as that from knee on movement

Post-op: questionnaire at 18 months; VAS pain at rest and with movement

Analysis: stepwise (forward) logistic regression analysis; 8 variables (dichotomised on median value); inclusion at was p=0.05

Pain with movement measured by Pain Matcher was omitted from the regression analysis because of a high drop-out rate (n=17) resulting from difficulties in comparing pain elicited by electrical device with that from joint […..]

Endpoint of follow-up: 18 months

 For how many participants were no complete outcome data available?

10%

Reasons for incomplete outcome data described?

No

Associations between pre-op predictors and VAS pain at rest and VAS pain during movement, at 18 months follow-up; logistic regression; OR [95% CI]

At 18 months post-op: 34% had no pain at rest or with movement; 24%

had pain at rest and 66% had pain with movement;  all patients with pain at rest also had pain with movement

Pain at rest (VAS)

Pre-op pain at rest (VAS)

6.48 [1.32;31.96]

Pre-op pain threshold (Matcher)

9.19 [1.69;50.07]

Other co-variates (Age, Duration of knee pain, Pain with movement [VAS], sensation threshold [Matcher], pain threshold [Matcher])  were statistically non-significant (p>0.05) [matched pain was omitted because of high drop-out]

The authors conclude that less pain relief was observed for patients with a high pre-operative VAS score for

pain at rest and a low pain threshold; both features which may reflect a central sensitisation mechanism.

Note: 23% of patients had no pain at rest. Patients with severe pre-op pain with movement had good pain outcomes . In contrast, patients with severe pre-op pain at rest benefited the least.

 ‘This seems to refute the concept that pain at rest should be a mandatory feature of those scheduled for TKR’

‘Thus, a distinction between pain at rest and pain with movement in pre-operative planning appears to

be of prognostic relevance.’

Note: NSAIDs were withdrawn between 3-14 days prior to surgery (: ‘routine practice’); some indications for selective outcome reporting

Singh 2008

Retrospective cohort

Hospital (Mayo Clinic); Rochester, USA

Project funding not stated; personal commercial funding [risk of bias unlikely]

Inclusion criteria: prim or rev TKA during 1996-2004; alive at 2- or 5-year followup; absence

of revision for primary TKA or rerevision for revision TKA; response to pain questionnaire at 2 or 5 years

8182 in database

2-year followup: 5290 of 7770 eligible patients

with prim TKA (68%)

5-year followup: 2602 of 4647 eligible patients with prim TKA (56%)

N=5290 (2 y followup)

N= 2602 (5y followup)

Mean age

68 years

Sex

44% Male

Diagnosis

OA 94%

RA 3%

Pre-op: medical records (Mayo registry), gender, age, and pre-op

pain severity

Pain question similar to that in AKS (no pain, mild (occasional), stairs only, walking and

stairs, moderate (occasional), moderate (continuous),and severe).

Post-op: routine clinical followup and mail at 2 and 5 years; post-op pain severity (dichotomized: moderate/severe versus no/mild (stairs only / walking and stairs, set to missing; 5% of total responses)

Analysis: of post-op moderate-severe pain by multivariate logistic regression models that adjusted for gender, age, and preop pain simultaneously [Note: strategy for choosing covariates not stated]

Endpoint of follow-up:  2 years and 5 years post-op

For how many participants were no complete outcome data available?

32% (2 y followup)

44% (5 y follow-up)

Reasons for incomplete outcome data described?

No (except death; non-responders)

Associations between pre-op predictors and moderate-severe pain, at 2 year and 5 year follow-up; logistic regression; OR [95% CI]

Only data for prim TKA patients are shown

Outcome: based on question similar to pain question in AKS; dichotomized (moderate/severe versus no/mild)

Mod-severe pain at 2 yr follow-up

OR [95% CI]; 6.6% male and 9% female patients had mod-severe pain

Age

<60 y (ref)

60–70 y 0.49 [0.31;0.77]

70–80 y 0.68 [0.44;1.05]

>80 y 1.18 [0.65;2.13]

Gender

Male (ref)

Female 1.45 [1.01;2.08]

Pre-op pain

Mild/none (ref)

Mod/severe 1.31 [0.57;3.03]

Mod-severe pain at 5 year follow-up

OR [95% CI]; 6.5% male and 7.9% female patients had mod-severe pain

Age

<60 y (ref)

60–70 y 0.66 [0.35;1.25]

70–80 y 0.73 [0.38;1.40]

>80 y 0.70 [0.21;2.31]

Gender

Male (ref)

Female 1.23 [0.74;2.02]

Pre-op pain

Mild/none (ref)

Mod/severe 4.60 [0.61;34.74]

Authors conclude that female gender and younger age predict greater risk of moderate-severe pain postoperatively in patients with primary TKA.

Note: associations (ORs) are statistically significant at 2 years follow-up, but not at 5 years follow-up

Note: relatively high loss to follow-up (in particular at 5 years); only limited adjustment for confounders; method of selecting confounders not described

‘Nonresponders were more likely to be female and to be 60 years or younger and both factors were associated with worse outcome’

Lingard 2007

Prospective cohort

International, multicentre study; 13 centres from USA, UK, Australia, Canada

Commercial funding (Stryker); some authors received benefits [but risk of bias is low]

Inclusion criteria: prim TKA with Kinemax prosthesis (Stryker); OA

Sept 1997 - Dec 1999

Exclusion criteria: history knee joint infection; previous implant on index knee; inability to complete questionnaires; bilateral TKA within 12 months

total 1249 (79%) eligible patients; exclusion 275 patients for bilat TKA

974 patients included; 22 w/o complete preop MH-score excluded from analyses

N= 952 patients

172 (18%) had preop MH score < age and gender-based median

= distressed group

780 (82%) had score > median

= nondistressed group

Distress / non-distress

Age

70 / 71 (p=0.03)

Sex (%Male)

34% 41% (p=0.11)

Income (%low)

30% / 23% (p=0.07)

Marital (separated)

15% / 6% (p<0.0001)

Lives alone

38% / 30% (p=0.04)

BMI

29.5 / 28.4 (p=0.004)

Comorb (no.)

2 (1-3) / 1 (1-2) p=0.004

Self-rep depress

25% / 5% (p<0.0001)

SF-36 MH (points)

Preop. 44 / 80 (p<0.0001)

Pre-op: Self-Admin Comorb questionnaire (validated); demographic details, socioecon data , comorbid conditions; SF-36 MH (mental health)

Note: SF-36 MCS not chosen because it can be influenced a patient’s physical health

Post-op: WOMAC pain, function (transformed so that higher=better; 100=best); SF-36 MH scale

Analysis: multivar analysis with preop MH (continuous) and WOMAC pain / function scores at follow-up as outcome; repeated measures analysis adjusted for age, gender, number of comorb conditions, country, center within country, and preop score.

in addition, multivar analyses after dichotomization (around age and gender-based median MH): distressed / non-distressed; adjusted (see above)

several sensitivity analyses using other cut-points

Endpoint of follow-up:  3, 12, 24 months post-op

For how many participants were no complete outcome data available?

28% (at 24 months = 2 years)

Reasons for incomplete outcome data described?

Yes (deaths; medical condition; revision surgery; withdrawal; ..)

Missing values imputed (LVCF-method); data also presented w/o imputation

Associations between pre-op MH score and post-op WOMAC pain / function at 3, 12 and 24 months ; absolute values [95% CI]; mean difference (p-value)

Only analyses after imputation of missing values are shown here

WOMAC-pain

Pre-op MH as continuous var

Regression coeff MH

12 months

 0.128 p<0.001

24 months

0.096 p=0.01

Each point increase in MH, increases WOMAC pain by 0.128 (12 months fup) or 0.096 (24 months fup); ‘better mental health associates with sign higher WOMAC pain scores indicating less pain’

Pre-MH after dichotomization:

Distressed versus non-distressed

12-months

72.9 [66.2;79.6]

77.6 [71.2;84.0] p=0.02

24-months

75.4 [68.7;82.0]

79.5 [73.2;85.8] p=0.03

WOMAC-function

Pre-op MH as continuous var

Regression coeff MH

12 months

 0.150 p=0.0001

24 months

0.106 p=0.007

Pre-MH after dichotomization:

Distressed versus non-distressed

12-months

64.1 [57.2;71.0]

67.0 [60.4;73.5] p=0.15

24-months

67.3 [60.4;74.1]

68.9 [62.4;75.5] p=0.4

Change in WOMAC-pain

Relative tot baseline

Distressed minus non-distressed

12 months

1.6 [−3.3;6.5] p=0.53

24 months

1.9 [−2.9;6.8] p=0.44

Change in WOMAC-function

Relative tot baseline

Distressed minus non-distressed

12 months

3.4 [−1.1;7.8] p=0.14

24 months

4.1 [−0.2;8.5] p=0.06

Authors conclude that many patients with psychological distress demonstrate a substantial decrease in that distress following surgery. Patients who are distressed have slightly worse pain preoperatively and for up to two years following knee arthroplasty as compared with patients with no psychological distress. With the exception of preoperative scores, these differences are not likely to be measurable at the individual patient level. WOMAC pain and function change scores do not differ between patients with and without distress after adjustment for covariates.

‘the magnitude of the effect of distress on pain and function appears to be very small’

‘suggests that preoperative distress is reversible, at least to some extent, and that this reversal is probably related to the surgical intervention’

Note: mental health of patients with pre- operative distress markedly improves following surgery, but that of patients classified preoperatively as non-distressed stays essentially the same

Note: clinically important differences between groups of patients are estimated to be on the order of 3 to 6 WOMAC points

Solomon 2006

Retrospective cohort; Medicare registry

Medicare data from 4 US states in year 2000

Non-commercial funding; conflicts of interest not stated

Inclusion criteria: all Medicare beneficiaries in 4 US states; primary TKR in 2000 (based on diagnosis in Medicare data)

411 hospitals Medicare/year 2000

6 merged/closed

295/405 (73%) response to survey

286 had Medicare patients

276 were confident of responded to survey

only hospitals where administrators answered “somewhat” or “very confident” included

N=276 hospitals

N= 9,073 patients

Age

74.2

Sex

31% Male

Indication TKA

OA 94.8%

Number comorbid

0 63.3%

1 23.2%

>1 13.5%

Pre-op: patient factors such as age, sex, Medicaid eligibility, number of comorbid conditions, and indication for TKR; hospital volumes (based on Medicare data per hospital); hospital characteristics (based on survey mailed to nurse administrators in 2002 responsible for orthopaedic surgery at each hospital)

Survey developed by multidisciplinary group with questions on preoperative evaluation, intraoperative care, and postoperative management; and question “how confident are you that information reported in this survey is accurate?”

Post-op: postoperative adverse events in first 90 days following surgery (composite of death, deep wound infection, pulmonary embolus, pneumonia requiring hospitalization, and acute myocardial infarction); based on Medicare claims

Analysis: stepwise selection and logistic multivariate regression model at 5% sign level; hospital volume was dichotomized (30% of cases and 5% of hospitals fell into low-volume category).

Developed summative index with variables from final multivariate model (‘preliminary index’)

Endpoint of follow-up:  90 days post-op

For how many participants were no complete outcome data available? Unclear

411 hospitals Medicare/year 2000

6 merged/closed

295/405 (73%) response to survey

286 had Medicare patients

276 were confident of resp to surv

Reasons for incomplete outcome data described?

No

Associations between predictors (patient and hospital characteristics) and peri-operative adverse events (composite) within 90 days; adjusted OR [95% CI]

Adverse events

Pulm embolus 0.9%

Acute MI 0.9%

Pneumonia 1.5%

Deep infect 0.3%

Death 0.6%

Total 3.6%

Predictors:

Comorbidity (number)

0 OR=1.0 (ref)

1 OR=1.5 [1.2;1.9]

2+ OR=1.7 [1.2;2.3]

Age

65-70 OR=1.0 (ref)

71-80 OR=1.3 [1.0;1.6]

81-95 OR=1.6 [1.1;2.4]

Sex

Male OR=1.6 [1.3;2.1]

Hospital volume

High (23+) OR=1.0 (ref)

Low (1-22) OR=1.6 [1.1;2.5]

Hospital survey scores

-pre-operative teaching program

No OR=1.8 [1.2;2.6]

-dedicated orthop operating room

No OR=1.3 [1.0;1.7]

Authors conclude that characteristics of hospital care, procedure volume, and patient-level factors are all associated with perioperative outcomes of TKR

 ‘our data suggest that volume does not explain much of the variation in outcomes, and other hospital characteristics are independent predictors of adverse events’

Note: number of comorbidities, older age, male sex, low hospital volume, the absence of a pre-operative training program, and not using a dedicated orthop operating room, independently associated with an increased risk of peri-operative adverse events

Note: hospital survey conducted 2-3 years after procedures had been performed (risk of recall bias); 27% of hospitals did not respond to survey and were excluded