Patellofemoral Arthroplasty Results in Better Time-weighted Patient-reported Outcomes After 6 Years than TKA: A Randomized Controlled Trial

Abstract

Background

In a previous study, we reported the 2-year outcomes of a parallel-group, equivalence, randomized controlled trial (RCT; blinded for the first year) comparing patellofemoral arthroplasty (PFA) and TKA for isolated patellofemoral osteoarthritis (PF-OA). We found advantages of PFA over TKA for ROM and various aspects of knee-related quality of life (QOL) as assessed by patient-reported outcomes (PROs). Register data show increases in PFA revision rates from 2 to 6 years after surgery at a time when annual TKA revision rates are decreasing, which suggests rapidly deteriorating knee function in patients who have undergone PFA. We intended to examine whether the early advantages of PFA over TKA have deteriorated in our RCT and whether revision rates differ between the implant types in our study after 6 years of follow-up.

Questions/purposes

(1) Does PRO improvement during the first 6 postoperative years differ between patients who have undergone PFA and TKA? (2) Does the PRO improvement at 3, 4, 5, and 6 years differ between patients who have undergone PFA and TKA? (3) Do patients who have undergone PFA have a better ROM after 5 years than patients who have had TKA? (4) Does PFA result in more revisions or reoperations than TKA during the first 6 postoperative years?

Methods

We considered patients who had debilitating symptoms and PF-OA as eligible for this randomized trial. Screening initially identified 204 patients as potentially eligible; 7% (15) were found not to have sufficient symptoms, 21% (43) did not have isolated PF-OA, 21% (43) declined participation, and 1% (3) were not included after the target number of 100 patients had been reached. The included 100 patients were randomized 1:1 to PFA or TKA between 2007 and 2014. Of these, 9% (9 of 100) were lost before the 6-year follow-up; there were 12% (6 of 50) and 0% (0 of 50) deaths (p = 0.02) in the PFA and TKA groups, respectively, but no deaths could be attributed to the knee condition. There were no differences in baseline parameters for patients who had PFA and TKA, such as the proportion of women in each group (78% [39 of 50] versus 76% [38 of 50]; p > 0.99), mean age (64 ± 9 years versus 65 ± 9 years; p = 0.81) or BMI (28.0 ± 4.7 kg/m² versus 27.8 ± 4.1 kg/m²; p = 0.83). Patients were seen for five clinical follow-up visits (the latest at 5 years) and completed 10 sets of questionnaires during the first 6 postoperative years. The primary outcome was SF-36 bodily pain. Other outcomes were reoperations, revisions, ROM, and PROs (SF-36 [eight dimensions, range 0 to 100 best, minimum clinically important difference {MCID} 6 to 7], Oxford Knee Score [OKS; one dimension, range 0 to 48 best, MCID 5], and Knee Injury and Osteoarthritis Outcome Score [KOOS; five dimensions, range 0 to 100 best, MCID 8 to 10]). Average PRO improvements over the 6 years were determined by calculating the area under the curve and dividing by the observation time, thereby obtaining a time-weighted average over the entire postoperative period. PRO improvements at individual postoperative times were compared for the patients who had PFA and TKA using paired t-tests. Range of movement changes from baseline were compared using paired t-tests. Reoperation and revision rates were compared for the two randomization groups using competing risk analysis.

Results

Patients who underwent PFA had a larger improvement in the SF-36 bodily pain score during the first 6 years than those who underwent TKA (35 ± 19 vs. 23 ± 17; mean difference 12 [95% CI 4 to 20]; p = 0.004), and the same was true for SF-36 physical functioning (mean difference 11 [95% CI 3 to 18]; p = 0.008), KOOS Symptoms (mean difference 12 [95% CI 5 to 20]; p = 0.002), KOOS Sport/recreation (mean difference 8 [95% CI 0 to 17]; p = 0.048), and OKS (mean difference 5 [95% CI 2 to 8]; p = 0.002). No PRO dimension had an improvement in favor of TKA. At the 6-year time point, only the SF-36 vitality score differed between the groups being in favor of PFA (17 ± 19 versus 8 ± 21; mean difference 9 [95% CI 0 to 18]; p = 0.04), whereas other PRO measures did not differ between the groups. At 5 years, ROM had decreased less from baseline for patients who underwent PFA than those who had TKA (-4° ± 14° versus -11° ± 13°; mean difference 7° [95% CI 1° to 13°]; p = 0.02), but the clinical importance of this is unknown. Revision rates did not differ between patients who had PFA and TKA at 6 years with competing risk estimates of 0.10 (95% CI 0.04 to 0.20) and 0.04 (95% CI 0.01 to 0.12; p = 0.24), respectively, and also reoperation rates were no different at 0.10 (95% CI 0.04 to 0.20) and 0.12 (95% CI 0.05 to 0.23; p = 0.71), respectively.

Conclusion

Our RCT results show that the 2-year outcomes did not deteriorate during the subsequent 4 years. Patients who underwent PFA had a better QOL throughout the postoperative years based on several of the knee-specific outcome instruments. When evaluated by the 6-year observations alone and without considering earlier observations, we found no consistent difference for any outcome instruments, although SF-36 vitality was better for patients who underwent PFA. These combined findings show that the early advantages of PFA determined the results by 6 years. Our findings cannot explain the rapid deterioration of results implied by the high revision rates observed in implant registers, and it is necessary to question indications for the primary procedure and subsequent revision when PFA is in general use. Our data do not suggest that there is an inherent problem with the PFA implant type as otherwise suggested by registries. The long-term balance of advantages will be determined by the long-term QOL, but based on the first 6 postoperative years and ROM, PFA is still the preferable option for severe isolated PF-OA. A possible high revision rate in the PFA group beyond 6 years may outweigh the early advantage of PFA, but only detailed analyses of long-term studies can confirm this.

Level of Evidence

Level I, therapeutic study.

Introduction

In severe cases, isolated patellofemoral osteoarthritis (PF-OA) may be treated surgically with patellofemoral arthroplasty (PFA) or TKA. The availability of the two options has initiated a debate of which is better—and how to define better. National arthroplasty registries have demonstrated PFA revision rates far greater than TKA revision rates during the first few years [24, 27]. A study from seven implant registers found 2-year revision rates in the range 2% to 9% for PFA, while the 6-year revision rates had risen to 15% to 24% [16]. The most common PFA failure modes are osteoarthritis progression in the nonreplaced tibiofemoral compartments, pain, and aseptic loosening [5, 35]. The continuing rise in revision rates from 2 to 6 years happens at a time when annual TKA revision rates are decreasing. This divergence suggests an increase in pain and discomfort and a decline in knee function beyond 2 years for patients who underwent PFA. It has been argued that revision of a PFA to a TKA should not be considered a failure of the index procedure [21], and others have suggested that the high revision rates recorded by registers are caused by multiple factors and not necessarily inherent in the implant concept [20, 22]. However, one can hardly defend a high revision rate for a particular implant type unless clear advantages outweigh the revision rate, or it can be shown that the observed revision rate is mainly the result of confounding and bias.

To understand and appropriately interpret the revision rates of PFA and TKA observed by registries, a link between the high PFA revision rate and deteriorating knee function (including pain) is essential. Quality of life (QOL) measures assess the patient’s subjective function and feelings, and QOL is the one dimension that may trump revision rates [6, 14, 21]. We believe that patients’ QOL should be the main concern whenever an operation is contemplated. We also believe that QOL should be quantified by summing up the QOL during the patients’ entire postoperative lifespan rather than looking at any single point in time. Cohort and case-control studies have suggested that PFA provides functional results and QOL as measured by patient-reported outcomes (PROs) on par with TKA, and that PFA results last beyond the first few years [6, 17]. These results reveal a paradox in the PFA versus TKA debate: PFA yields a good and lasting QOL and yet has a higher revision rate. Variation in the indication for the primary PFA procedure and for subsequent revisions are difficult to control for in statistical analyses of register data, and the variation may well result in strong confounding and bias. The only way to compare the inherent properties of PFA and TKA implants for treating PF-OA is to perform a study, which by randomization cancels out other differences between PFA and TKA patients, and that allows QOL results to be studied over time rather than the crude surrogate measure of revision rate. We initiated such a randomized controlled trial (RCT) 14 years ago [11, 28], and since all the patients have now crossed the 6-year follow-up point, we may determine whether a rapid decline of outcomes within the first 6 years is seen in the PFA group compared with the TKA group as suggested by registry data.

We therefore reported on the midterm results from our RCT, defined as the results after 6 years of follow-up. The study intended to answer the basic question of which implant produces better results. The primary outcome was QOL, which we measured with the SF-36 bodily pain measure, and the secondary outcomes were other PROs and knee ROM. There are two important aspects to the main question: the average QOL during the entire postoperative period and the QOL at specific postoperative times.

We therefore asked the following: (1) Does PRO improvement during the first 6 postoperative years differ between patients who have undergone PFA and TKA? (2) Does the PRO improvement at 3, 4, 5, and 6 years differ between patients who have undergone PFA and TKA? (3) Do patients who have undergone PFA have a better ROM after 5 years than patients who have had TKA? (4) Does PFA result in more revisions or reoperations than TKA during the first 6 postoperative years?

Patients and Methods

Study Overview

The study was a multicenter, parallel-group, randomized controlled equivalence trial of 100 patients [28], with the goal of determining whether PFA was therapeutically similar to TKA. The sample size calculation was based on SF-36 bodily pain, and it indicated that 40 patients would be needed in each group [2]. The inclusion of 100 patients was meant to allow for up to 10 withdrawals in each randomization group during the planned study duration of 10 years, but the design also meant that the power would not be sufficient for subgroup analyses such as sex. Patients were eligible if they had severe symptoms of PF-OA that had not responded to nonoperative treatment and had bone-on-bone contact visible on the tangential radiograph with preserved tibiofemoral joint lines on the frontal radiograph. The exclusion criteria were inability to provide informed consent, non-Danish citizenship, disseminated malignant disease, severe neurologic conditions, rheumatoid arthritis, knee malalignment, patella alta/infera, limited ROM, and complex regional pain syndrome (in case of previous procedure to knee). The trial compared PFA (Avon, Stryker) and TKA (fixed bearing, cruciate retaining, PFC Sigma, Johnson and Johnson DePuy) for the treatment of PF-OA [28]. The randomized 1:1 allocation to the PFA or TKA group was done intraoperatively after assessing the joint pathology (full-thickness tibiofemoral articular cartilage lesions with a diameter of more than 6 mm led to exclusion) using sealed, randomly ordered and numbered envelopes. The PFA surgical technique has been described elsewhere [1, 26]. The operations were performed in seven participating hospitals from June 2007 through October 2014, and all patients have currently passed the 6-year follow-up date. Patients, all non-OR staff, physiotherapists, and general practitioners were blinded to the implant, and hospital notes and discharge letters were modified to conceal implant information. The primary outcome was the SF-36 bodily pain score. The secondary outcomes were other PROs (see Outcome Measures), ROM, and revisions and reoperations. The trial was recorded in ClinicalTrials.gov (NCT01326156).

Patients

Based on the initial screening of patients referred to the participating centers, 204 patients were identified as potentially eligible. Of these, 7% (15) were found not to have sufficient symptoms, 21% (43) did not have isolated PF-OA, 21% (43) declined participation, 1% (3) were not included after the target number of 100 patients had been reached (Supplementary Fig. 1; http://links.lww.com/CORR/A756). There were no differences in baseline variables for patients who underwent PFA and TKA, such as the proportion of women in each group (78% [39 of 50] versus 76% [38 of 50]; p > 0.99), mean age (64 ± 9 years versus 65 ± 9; p = 0.81) or BMI (28.0 ± 4.7 versus 27.8 ± 4.1 kg/m²; p = 0.83) (Supplementary Table 1; http://links.lww.com/CORR/A757).

Because of withdrawals from the study, the number of available patients gradually decreased (Supplementary Fig. 1; http://links.lww.com/CORR/A756). At all planned postoperative contact points, some patients did not complete the questionnaire or did not attend clinical follow-up visits. These patients were traced, and some patients had severe disease or had died. Some patients completed paper questionnaires that were lost at the included hospitals before being coded and stored in the research database. This resulted in 91% to 100% data completeness for individual follow-up contact points (Table 1).

Table 1.

Data completeness, patient availability, and withdrawals

	Withdrawals			Missed data points		For analysis
Event	Death	Other reasons	Available	Declined	Lost	Number	% of available
Inclusion			100			100	100
Preassessment			100			100	100
Baseline PROM			100		7	93	93
Operation			100			100	100
2-week follow-up			100	6		94	94
6-week PROM			100	6		94	94
3-month PROM			100	2		98	98
4-month follow-up			100	3		97	97
6-month PROM			100	6		94	94
9-month PROM			100	5		95	95
1-year PROM	1		99	1		98	99
1-year follow-up	1		99	2		97	98
2-year PROM	2	1	97	4		93	96
2-year follow-up	2	1	97	4		93	96
3-year PROM	2	2	96	5		91	95
4-year PROM	2	2	96	9		87	92
5-year PROM	3	3	94	4		90	96
5-year follow-up	3	3	94	8		86	91
6-year PROM	5	4	91	3		88	97

Deaths
Sex	Age at inclusion	Event time	Cause of death	Group	Related to knee
Male	66	11 months	Suicide	PFA	No
Female	73	1 year, 6 months	Cancer	PFA	No
Male	65	2 years, 2 months	Infection, sepsis	PFA	No
Female	71	4 years 2 months	Cancer	PFA	No
Male	71	5 years	Cardiac arrest	PFA	No
Female	73	5 years 4 months	Meningitis	PFA	No

Withdrawals, other reasons
Sex	Age at inclusion	Event time	Reason	Group	Comments
Male	65	1 years, 11 months	Severe hematologic disease	PFA	Died 2 years, 2 months postoperatively
Male	61	2 years, 9 months	Cerebral hemorrhage	TKA
Male	79	4 years, 11 months	Mb. Alzheimer	TKA	Died 6 years, 2 months postoperatively
Female	66	5 years 1 months	Unhappy with questionnaire	PFA

The patient who withdrew before the 2-year follow-up died between 2 and 3 years postoperatively, and this explains why the total number of deaths in this table of withdrawals is less than the number of deaths shown in the CONSORT diagram (Supplementary Fig. 1; http://links.lww.com/CORR/A756); PROM = patient-reported outcome measure.

Of the 100 randomized patients, 12% (6 of 50) of patients in the PFA group and 0% (0 of 50) of patients in the TKA group died within 6 years of the index operation (p = 0.02). No deaths could be attributed to the knee condition (Table 1).

Outcome Measures

Patients completed PRO questionnaires preoperatively and at 6 weeks, 3 months, 6 months, 9 months, 12 months, and yearly thereafter. Of relevance to the present study, the questionnaire included the SF-36 [36], Oxford Knee Score (OKS) [7], and Knee Injury and Osteoarthritis Outcome Score (KOOS) [32]. The SF-36 has eight dimensions, each with a range from 0 to 100 (best score), and the MCID is 6 to 7 [2, 10]. The OKS is one-dimensional with a score ranging from 0 to 48 (best score), and the MCID is about 5 [4]. The KOOS has five dimensions with scores ranging from 0 to 100 (best score), and the MCID is 8 to 10 [31].

Revisions and Reoperations

We retrieved information about revisions and reoperations from patients and through the Danish National Patient Register, in which all public and private health encounters are registered (ICD-10 diagnoses and Nordic Medico-Statistical Committee (NOMESCO) procedure codes) with the unique Danish person identifier as key. We searched the Danish National Patient Register for all knee procedures and femoral amputations, and details were retrieved from hospital records. The Danish Civil Registration System was searched for all deaths, and we determined the causes of death using hospital records. Danish public health records have very high completeness, and it is very unlikely that a single event would have been missed with this process.

Research Questions

To answer the first research question about postoperative average improvement of PROs, we calculated the area under the curve (AUC) for the primary outcome SF-36 bodily pain change from baseline and divided the AUC by the observation time (72 months), reaching a time-weighted mean change [19, 28, 34]. We did the same for the secondary outcomes (other SF-36 dimensions, KOOS dimensions, and OKS). The average (time-weighted) change from baseline is a maximum-likelihood estimate of a random patient’s change from baseline at a random time during the first 6 years.

To answer the second question about QOL at specific postoperative times, we analyzed PRO changes from baseline at individual postoperative timepoints. So instead of determining the average improvement over the entire postoperative period (first question), we specifically looked at the improvement at 3, 4, 5, and 6 years postoperatively for the two implant types, and we considered both knee-specific (KOOS and OKS) and generic (SF-36) QOL.

To answer the third question about ROM, we analyzed the change in ROM from baseline to the 5-year follow-up.

To answer the fourth question about reoperations and revision, we described subsequent procedures and performed competing risk analyses on revisions and other reoperations. A traditional Kaplan-Meier analysis would produce inaccurate estimates because the chance of observing an event (for example, reoperation) is affected by the occurrence of competing events (such as revision or death) [15].

Ethical Approval

Ethical approval was obtained from the Ethical Committee System in Denmark, Region Midtjylland, reference numbers 20070025 and 1-10-72-195-21.

Statistical Analyses

Ordinal data are described using frequencies, and distributions were compared using the Fisher exact test. Continuous data are described as the mean ± SD and were tested with the Welch t-test.

All analyses of changes were based on paired differences between given timepoints for individual patients. The area under the curve for PRO and ROM changes from baseline was calculated [19, 28].

We analyzed the time-dependent functions that describe revisions and reoperations (other than revision) using the competing risk method, and the difference between group estimates was tested using the Gray test. We performed two competing risk analyses. The first focused on reoperations other than revision, with death and revision as competing events. The second analysis focused on revision with death as a competing event.

We performed all analyses using the intention-to-treat principle; thus, patients allocated to a randomization group continued to be analyzed as such, irrespective of possible revision. Significance was set at 5%. All SF-36 calculations were based on QualityMetric Health Outcomes Scoring Software, Version 5.0.6163.22119. All other analyses were performed using the R statistical computing and graphics environment, version 3.6.3 [30].

Results

Average PRO Improvement During the First 6 Postoperative Years

During the first 6 years, patients who underwent PFA demonstrated consistently better knee-related PROs than patients who underwent TKA. The mean improvement from baseline of the primary outcome, SF-36 bodily pain, was larger during the first 6 postoperative years for PFA patients (35 ± 19) than for TKA patients (23 ± 17, mean difference 12 [95% CI 4 to 20]; p = 0.004) (Table 2). The change from baseline during the first 6 postoperative years was also in favor of PFA for secondary outcomes SF-36 physical functioning (mean difference 11 [95% CI 3 to 18]; p = 0.008), KOOS Symptoms (mean difference 12 [95% CI 5 to 20]; p = 0.002), KOOS sport/recreation (mean difference 8 [95% CI 0 to 17]; p = 0.048), and OKS (mean difference = 5 [95% CI 2 to 8]; p = 0.002).

Table 2.

Mean change during the first 6 years from baseline of the 14 PRO dimensions studied

Patient-reported outcome	PFA time-adjusted change a	TKA time-adjusted change a	Mean difference (95% CI)	p value
SF-36 Bodily pain	35 ± 19	23 ± 17	12 (4 to 20)	0.004
SF-36 Physical role	25 ± 33	23 ± 39	2 (-13 to 18)	0.79
SF-36 Physical functioning	25 ± 18	14 ± 18	11 (3 to 18)	0.008
SF-36 Emotional role	18 ± 40	14 ± 43	4 (-14 to 22)	0.65
SF-36 Vitality	12 ± 19	7 ± 18	5 (-3 to 13)	0.18
SF-36 Social role	6 ± 17	2 ± 20	4 (-4 to 11)	0.38
SF-36 Mental health	6 ± 14	2 ± 15	5 (-2 to 11)	0.15
SF-36 General health	4 ± 16	4 ± 15	1 (-6 to 7)	0.88
KOOS Pain	30 ± 15	24 ± 18	6 (-1 to 13)	0.08
KOOS Symptoms	20 ± 17	8 ± 18	12 (5 to 20)	0.002
KOOS ADL	25 ± 14	24 ± 17	2 (-5 to 8)	0.60
KOOS Sport/Recreation	20 ± 20	11 ± 19	8 (0 to 17)	0.048
KOOS QoL	31 ± 18	26 ± 19	6 (-2 to 14)	0.15
OKS	13 ± 6	9 ± 7	5 (2 to 8)	0.002

^aData are presented as mean ± SD; SF-36 range is 0-100 (best), MCID is 6-7 for physical dimensions; KOOS range is 0-100 (best), MCID is 8-10; OKS range is 0-48 (best), MCID is 5.

All 14 PRO dimensions showed improvement in favor of PFA, but only the just mentioned differences were statistically significant and of a magnitude that suggests clinical importance (Table 2). As a reference to interpret the differences between PFA and TKA, the minimum clinically important difference (MCID) for the SF-36 physical dimensions is 6 to 7 [2, 10], 8 to 10 for the KOOS [31], and approximately 5 for the OKS [4].

PRO Improvement at 3, 4, 5, and 6 Years

With regard to PRO improvements beyond 2 years, patients who underwent PFA demonstrated inconsistent advantages versus those who had TKA. At 3 years, the mean improvement from baseline of the primary outcome, SF-36 bodily pain score, was larger for patients who underwent PFA (41 ± 25) than for patients who underwent TKA (29 ± 25) (mean difference = 13 [95% CI 1 to 24]; p = 0.03); however, we observed no difference at later postoperative times up to 6 years (Fig. 1).

Fig. 1.

A-H These graphs show changes in the eight SF-36 dimensions from baseline: (A) bodily pain, (B) physical role functioning, (C) physical functioning, (D) emotional role functioning, (E) vitality, (F) social role functioning, (G) mental health, (H) general health perceptions. The p values at the top of each panel relate to the comparison of the two randomization groups at the individual postoperative times. The error bars denote the 95% CI of the mean (Supplementary Table 2; http://links.lww.com/CORR/A758). A color image accompanies the online version of this article.

The secondary SF-36 outcomes of physical functioning and vitality revealed differences between the implants in favor of PFA (Fig. 1). The SF-36 physical functioning was higher for PFA at 3 years (mean difference 13 [95% CI 2 to 23]; p = 0.03). The SF-36 vitality was higher for PFA at 6 years (mean difference 9 [95% CI 0 to 18]; p = 0.04) (Supplementary Table 2; http://links.lww.com/CORR/A758). All SF-36 dimensions stabilized after 6 weeks to 3 months for the PFA group and after 3 months for the TKA group (Supplementary Fig. 2; http://links.lww.com/CORR/A759). We observed no statistically significant or clinically important deteriorations from 2 to 6 years for any SF-36 dimensions.

From baseline to 3 years, patients who underwent PFA had more improvement in our secondary outcome, KOOS symptoms, compared with patients who had TKA (mean difference 10 [95% CI 1 to 19]; p = 0.038); no other KOOS differences were demonstrated at 3 years or beyond (Fig. 2). There was no improvement in the PFA group after 3 months, whereas improvement in the TKA group continued for 9 months to 1 year (Supplementary Fig. 3; http://links.lww.com/CORR/A760). No KOOS dimensions showed deterioration from 2 to 6 years.

Fig. 2.

A-E These graphs show changes in the five KOOS domains from baseline: (A) function in activities of daily living (ADL), (B) symptoms, (C) pain, (D) knee-related quality of life (QOL), (E) function in sport and recreation (sport/recreation). The p values at the top of the panels relate to the comparison of the PFA and TKA groups. The error bars denote the 95% CI of the mean (Supplementary Table 2; http://links.lww.com/CORR/A758). A color image accompanies the online version of this article.

There was no difference in the OKS improvement from baseline between the groups beyond 12 months (Fig. 3). OKS improvement in the PFA group stabilized after 6 weeks to 3 months, whereas this took 6 to 9 months for the TKA group. For both groups, there was no deterioration in the outcome at later times (Supplementary Fig. 4; http://links.lww.com/CORR/A761).

Fig. 3.

This graph shows changes in the OKS from baseline. The p values at the top of the panels relate to the comparison of the PFA and TKA groups. The error bars denote the 95% CI of the mean (Supplementary Table 2; http://links.lww.com/CORR/A758). A color image accompanies the online version of this article.

ROM

Both patients who underwent PFA and those who underwent TKA had lost some ROM from baseline to 5 years, but the loss of those in the PFA group was less than the loss of those in the TKA group. The preoperative ROM was 133° ± 11° and 132° ± 12° for the PFA and TKA groups, respectively. At 5 years, ROM was 130° ± 10° and 121° ± 10°, respectively. Throughout the observation period, patients undergoing PFA had a smaller decrease in ROM than those undergoing TKA, except for the 2-week follow-up (Fig. 4). At 5 years, the mean change in ROM from baseline for PFA and TKA was -4°± 14° (95% CI -8° to 1°; p = 0.12) and -11° ± 13° (95% CI -15° to -7°; p < 0.001) (mean difference 7° [95% CI 1° to 13°]; p = 0.02).

Fig. 4.

This graph shows changes in ROM from baseline. The decimal numbers between the red and blue curves are p values of the comparison of the PFA- and TKA-change. The p values at the top of the panel relate to comparisons of the PFA and TKA change from baseline. The error bars denote the 95% CI of the mean. A color image accompanies the online version of this article.

Reoperations and Revisions

At 6 years, there was no difference in the risk of revision or reoperation (other than revision) between the PFA and TKA patients. Five patients who had a PFA and six patients who underwent TKA had a reoperation other than revision within 6 years (Supplementary Table 3; http://links.lww.com/CORR/A762). The competing risk analysis on reoperations other than revision and Gray’s test for equality across groups showed no difference (PFA and TKA estimates at 72 months were 0.10 (standard error 0.04 [95% CI 0.04 to 0.20) and 0.12 (standard error 0.05 [95% CI 0.05 to 0.23; p = 0.71). Five patients who had a PFA and two patients who underwent TKA were revised within 6 years (Supplementary Table 4; http://links.lww.com/CORR/A763). The competing risk analysis on revision showed no difference (PFA and TKA estimates at 6 years were 0.10 [standard error 0.04 [95% CI 0.04 to 0.20] and 0.04 [standard error 0.03 [95% CI 0.01 to 0.12]; p = 0.24]).

Discussion

Our main aim for this RCT was to compare the outcomes of PFA and TKA to guide surgeons to the better treatment. We wanted to collect knee-specific and generic QOL measures using PROs, and we intended to collect these measures frequently enough to allow calculations of average improvements over the entire postoperative study period rather than focus on any particular follow-up time. In a previous publication from this RCT [28], we found that PFA has clear advantages over TKA up to 2 years. Since registers have shown increased revision rates from 2 to 6 years [16], the present study had the dual purpose of analyzing results up to 6 years and determining whether patients who underwent PFA still had an advantage over those who had a TKA or whether their results had deteriorated as implied by register publications [16, 24, 27]. The main findings of the analyses are that (1) patients who underwent PFA still have an advantage, since their average QOL during the 6 postoperative years is better than that of patients who underwent TKA as judged by several knee-related patient-reported outcome dimensions, and (2) those who underwent PFA have a better range of movement than patients who underwent TKA. We also found that an isolated analysis of the QOL at 6 years after the operation failed to show a difference between the two groups. Since the patients who underwent PFA have had a better QOL throughout the postoperative period and since their QOL at 6 years is no worse than that of those who underwent TKA, we suggest that PFA is used for patients with PF-OA subject to the limitations mentioned in the next paragraph.

Limitations

There are several limitations to the study, and the first relates to the PRO instruments used. The three PRO systems we included (SF-36, KOOS, and OKS) are meant to cover both generic and knee-specific QOL. There is, however, no gold standard PRO instrument for arthroplasty treatment of PF-OA. The KOOS was developed as an extension to the WOMAC for young- to middle-aged patients with a knee injury and/or osteoarthritis [32]. Its content validity for patients with PF-OA and other conditions is unknown [13]. The same limitation applies to the OKS instrument, which was developed for patients who underwent TKA, irrespective of pathology [7]. We used the OKS and KOOS systems as “generic knee-specific” systems, expecting them to reflect relevant aspects in patients with PF-OA. This usage of PRO instruments is common and acceptable for specific subpopulations, until PRO systems designed and validated for those subpopulations become available. It must, however, be remembered that aspects of the well-being of patients with PF-OA may not be covered by the used PRO instruments, and conversely, the content of these PRO instruments may not be important to patients with PF-OA. There is a need for high-quality development studies of PRO systems for patients with PF-OA who undergo arthroplasty.

The use of MCID for evaluating the average (time-weighted) PRO change is reasonable, since PRO data is already averaged over time. For the SF-36, a patient is asked to consider the past 4 weeks, and this ill-defined average is the basis for MCID calculations. Extending the period for which the average is determined and providing a rigorous definition of the average should by induction not fundamentally change the validity of MCID, but we do acknowledge that this has not been the subject of investigation.

The main indication for PFA is severe PF-OA with bone-on-bone contact on the tangential view without damage to the tibiofemoral joint; patient characteristics play a minimal role for the indication. This opinion was used as an eligibility criterion for the study, and the results cannot be extrapolated to other indications. Our results cannot be used as an argument for performing PFA in patients with only mild degenerative changes or for neglecting mild/moderate tibiofemoral changes in a younger patient to avoid TKA.

Patient allocation was random, and patients were blinded for the first year. This study methodology is a strength that is likely to level out confounders between the two groups, which is documented by the equal distribution of baseline measures (including sex, age, BMI). There is, however, always a risk of an unequal distribution of other confounders, and comorbidity might be one of these that could affect outcomes. The unequal distribution of deaths might be an indication of this, but it might also be a chance finding. The fact that none of the generic SF-36 dimensions differed between the groups at baseline does, however, suggest that the level of comorbidity was equal in the groups [28].

We studied specific implants not generic implants. The TKA (PFC Sigma) was designed more than 20 years ago [3], and it is possible that newer TKA designs may affect the PFA/TKA balance of results. We do, however, believe, that the PFC Sigma is a reasonable TKA representative for a comparison with PFA, and the postulated benefits of newer TKA designs are yet to be proven. Different PFA designs may have different outcomes [5]. The PFA that we used (Avon) is characterized by a broad distal intercondylar extension that has the form of a swallow’s tail, which allows distal positioning without impinging on the cruciate ligaments [1, 26]. Other so-called third-generation trochlear components do not have the same geometry, and there is no data to suggest how this may affect the PFA results.

All surgical procedures in this study were performed by dedicated knee arthroplasty surgeons who were trained to use the PFA implant [28]. A seminar with presentations on indications, discussions of the surgical technique, and live surgery was organized when launching the study. The designer surgeon provided initial training to the first author, who then supervised each surgeon in the next group of surgeons for at least three procedures. The goal was to minimize variation, so that indications, surgical technique, the perioperative pathway would cause minimal confounding, and the outcome should mainly reflect differences in the inherent qualities of the two implant types for treating PF-OA. Before the study, PFA had been used sporadically by some of the participating surgeons, whereas others had never performed PFA. We do not have data on how many PFA procedures a knee surgeon should perform to excel in the procedure, but we believe that an understanding of indications and basic principles of the surgical procedure [1, 26] and supervision during the first cases are more important than numbers.

Analyses of the differences between patients who underwent PFA and those who underwent TKA were done on whole groups, and no analyses were performed on subpopulations, for instance sex or age groups, as the number of patients in our study is insufficient. Hence, one should consider the possibility that subpopulations may exist, where the whole-group conclusions do not apply.

ROM measurements were not blinded for the implant beyond 1 year, and even before the 1-year follow-up, measurements were quasiblinded. If we were to redesign the study, we would use a validated patient-reported instrument for ROM assessment in addition to clinical measurements [23].

Average PRO Improvement During the First 6 Postoperative Years

During the first 6 postoperative years, patients undergoing PFA demonstrated consistently better knee-related average PROs than patients undergoing TKA. The difference in time-weighted improvement from baseline demonstrate that a random patient who had a PFA is likely to have a clinically important and better knee-related QOL than a random patient who underwent TKA at a random time during the first 6 postoperative years. The data show that PFA is superior to TKA in terms of SF-36 bodily pain (primary outcome) and physical functioning, KOOS symptoms and sport/recreation, and OKS scores. No PRO dimension showed an average improvement during the first 6 years in favor of TKA. We conclude that patients who underwent PFA had a clinically meaningful advantage compared with those who underwent TKA. From this perspective, it is not so important if comparisons at fixed timepoints do not show differences. Every day of postoperative life counts for patients. Cohort studies and systematic reviews have focused on outcomes at a specific time and have based conclusions on findings at, for instance, 5 years. We do not agree with this approach.

PRO Improvement at 3, 4, 5, and 6 Years

Patients who underwent PFA had inconsistently better QOL scores 3, 4, 5, and 6 years after surgery than patients who underwent TKA. At 3 years, SF-36 bodily pain and physical functioning as well as KOOS symptoms were better for PFA than for TKA patients, and at 6 years, SF-36 vitality was better for PFA. However, the general picture is that there is no difference in the knee-related QOL measures at follow-up times beyond 3 years. We observed no consistent change in any of the PRO dimensions from 1 year onward for patients who underwent either PFA or TKA. This agrees with a previous cohort study that found no deterioration from 1-year observations to final follow-up at an average of 10 years [6].

ROM

At 5 years, patients undergoing PFA demonstrated better ROM than patients undergoing TKA (patients were seen for clinical follow-up at 4 months, 1, 2, and 5 years (a 10-year follow-up is planned). Our data demonstrate that patients who underwent PFA had regained their preoperative ROM at 12 months, whereas patients who had a TKA did not reach the baseline ROM, even at 5 years. We are not aware of a study that estimates a MCID for knee ROM after arthroplasty. A study of patients who have had a stroke has shown the MCID to be 7° to 8° [12], and we assume that similar differences in ROM can be important for patients with PF-OA. Studies have demonstrated a positive correlation between ROM and knee-related PRO [18, 29] as well as a positive effect on functional ability and satisfaction [8]. The preoperative ROM was 132°, which is more than that reported for TKA [9, 33], which may explain the decreased postoperative ROM in patients who underwent TKA relative to baseline.

Revisions and Reoperations

At 6 years, there was no difference in the risk of revision or any other reoperation between the patient groups. The cumulative revision rates at 6 years were 10% and 4% for PFA and TKA, respectively. This suggests more revisions in the PFA group, but our study was underpowered to demonstrate this. A sample size calculation aimed at demonstrating a difference between the mentioned frequencies would require a much larger sample size. We found varying indications for reoperations, and there may be a pattern of more manipulations under anesthesia in the TKA group and more procedures for patellofemoral instability in the PFA group.

Conclusion

During the first 6 years, patients who underwent PFA have better knee function than patients who underwent TKA, but isolated assessments at times beyond 3 years did not demonstrate a difference in outcome. Patients who underwent PFA had a better ROM than those who had TKA at 5 years. Medium-term function at 6 years of patients who have undergone PFA and TKA for PF-OA is similar, but more importantly, the early advantage of PFA has not been lost at 6 years, and no deterioration of PFA or TKA results are seen from 2 to 6 years postoperatively. We did not find a difference in revision or reoperation rates between the two implant types. Consequently, the reason for the high PFA revision rates observed by registers may be caused by indications and usages that are not inherently related to the PFA implant. We expect that long-term analyses of our data will show more revisions in the PFA group. There are three arguments for this. First, there could be disease progression in the tibiofemoral compartments [25]. Second, surgeons may be reluctant to perform TKA revisions, and conversely, there may be a lower threshold for performing PFA revisions. Third, general perceptions about the success of a certain type of implant will affect clinical practice, and registries uniformly publish reports of poor survival of PFA compared with TKA, which affects expectancy, clinical practice and revision rates. In this setting, adjusting clinical practice based solely on registry data without considering why revisions occur could result in a bias against an appropriate treatment. Only randomized studies with long-term follow-up will be able to provide answers on the preferable treatment of isolated patellofemoral osteoarthritis.