Return to Sports and Physical Activity After Total and Unicondylar Knee Arthroplasty: A Systematic Review and Meta-Analysis

The PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) statement was used for this systematic review [22]. A research protocol for this review was agreed by all co-authors before starting the literature searches. The study protocol was published online at the PROSPERO International prospective register of systematic reviews (http://​www.​crd.​york.​ac.​uk/​PROSPERO/​) under registration number: CRD42014009370.

The electronic databases MEDLINE (biomedical literature) via PubMed, Embase via OvidSP, and SPORTDiscus (sports and sports medicine literature) via EBSCO were searched for relevant literature. Searches were performed up until January 5, 2015. In all three databases the following three categories of keywords (and related synonyms) were used to build a sensitive search strategy and to provide a systematic search: ‘knee arthroplasty’, ‘sports’ and ‘recovery of function’. In MEDLINE we strived to use medical subject headings (MeSH), otherwise we searched the title and/or abstract (tiab). Furthermore, search terms were truncated through the use of a * symbol in order to find all terms beginning with a specific word. Within each keywords category, the different synonyms were combined using the Boolean command OR, and categories were linked with the Boolean command AND. The exact details of the search strategy can be found in the electronic supplementary material, Appendix S1.

The first author (SW) selected suitable studies with the assistance of a medical student and input from a medical librarian of the Academic Medical Centre (AMC). Inclusion criteria were (1) knee OA patients who underwent total and/or unicondylar KA; who (2) were active in a sport before the surgery and intended to resume or intensify their sporting activity; and (3) that included an outcome measure of interest to the authors. The primary outcome was the percentage (and number) of patients to RTS (preferably described in terms of sports level, duration and frequency) and time to RTS. Secondary outcomes were specific PA outcomes measures, namely University of California, Los Angeles (UCLA) rating score, Tegner-Lysholm rating scale and Grimby score [23‐25].

The reference lists of selected articles were screened to identify additional articles to be included. We also performed a forward search using ‘Web of Science’ to see which of these papers were referred to by other papers after they had been published.

We assessed the risk of bias of the studies using the Quality in Prognosis Studies (QUIPS) tool [26]. This quality assessment method considers six domains of potential biases: (1) study population; (2) study attribution; (3) prognostic factor information; (4) measurement of and controlling of confounding variables; (5) measurement of outcomes; and (6) analysis approaches. The first author (SW) assessed the quality of all selected articles, and this was repeated by two other authors (VG and PK), who each assessed risk of bias of 50 % of the selected articles. We customised the tool to our review by defining the issues of the domains to be scored. The details of these issues can be found in the electronic supplementary material, Appendix S2. In domain 2, we adjusted for a minimum follow-up period of 1 year, according to the literature, which states that the greater part of the knee function will have been regained at 1 year after surgery [27‐29].

By assessing response rate and information about non-responders, we chose a cut-off point of 20 %, based on previous studies in this field [30, 31]. In domain 5, concerning study confounding, we identified confounding variables for activity from previous research we found on this subject before performing this systematic review [1, 28, 32‐35]. We rated the issues per domain separately as ‘yes’, ‘no’, ‘partial’ or ‘unsure’, which then led to a risk of bias for each domain to being ‘low’, ‘moderate’ or ‘high’. We considered a study to have an overall low risk of bias when the methodological risk of bias was rated as low or moderate in all six domains, with at least four domains being rated ‘low’. A study was rated as having an overall high risk of bias if two or more of the domains scored ‘high’. In-between quality was scored as ‘moderate’.

The first author (SW) extracted data from all selected original articles, and this was repeated by two other authors (VG and PK), each extracting data from 50 % of the included articles.

All authors used a standardised data extraction form including the following topics: (1) study information: author, year, country and reference number; (2) study design and follow-up; (3) information about study population: cohort, population size, sex, age, body mass index (BMI), comorbidities, and type of OA (primary or secondary causes, such as systematic inflammatory disease or post-traumatic arthritis); (4) description of rehabilitation protocols used; (5) definition of the outcome measures; (6) preoperative activity and definition (e.g. presymptomatic or at time of surgery); (7) postoperative activity; (8) RTS percentages and time to RTS; and (9) confounding factors taken into account for RTS, such as sex, BMI, restricting comorbidities, complications, preoperative sports level, surgeon recommendations or other psychosocial influencing factors.

From the studies that described pre- and/or postoperative participation in specific types of sports and/or times to RTS, data were pooled and categorised into low-, intermediate- or high-impact sports, according to the levels of impact on the knee joint (see electronic supplementary material, Appendix S3). This classification is in compliance with Vail and Mallon [36] and supported by a biomechanical study from Kuster et al. [37], in which both peak loads and flexion angles of the knee were considered. We calculated pooled RTS percentages by comparing pooled pre- and postoperative sports participation data.

We retrieved a total of 1115 potentially relevant citations from our search. After deleting 286 duplicates and applying the inclusion criteria to titles and abstracts, we reviewed 37 full-text articles, 12 of which were review articles. We excluded these from data extraction, as with 14 other articles, which were excluded for various reasons such as current concept reviews, case reports or studies not presenting outcomes of our interest. On both reviews and the included 11 articles, we performed reference screening and forward citation tracking, which resulted in seven additional articles being included. The article by Jahromi et al. [38] was excluded because the same UKA cohort was described in the article by Walton et al. [39]. Finally, 18 original studies were included. The PRISMA flowchart of our search procedure can be found in Fig. 1.

All of the included studies were observational, 13 being cross-sectional studies, three prospective studies and two retrospective cohort studies. Two studies were performed in Australia [39, 40], one in Austria [41], two in France [42, 43], four in Germany [44‐47], one in Italy [48], one in Korea [49], one in Switzerland [14], four in the UK [50‐53] and two in the US [54, 55]. From three of the 18 included articles, data about sports activities after both TKA and UKA was able to be extracted, from ten after TKA and from five after UKA, of which one article specifically described outcomes after lateral UKA. Of the 13 articles with respect to data about RTS after TKA, the study population of eight studies was a non-selected group of KA patients. Five studies examined a selected population of patients. Two of these studies examined patients younger than 75 years old, one study assessed patients younger than 65 years old, another study concerned non-revised patients younger than 55 years old and the last study evaluated licensed judokas (i.e. people who participate in judo, which is a method of defending oneself or fighting without the use of weapons, based on jujitsu) with black belts of 60 years and older.

This latter study described outcomes of two different age groups, namely patients younger than 55 years and patients aged 65–75 years. Of the eight articles with respect to data about RTS after UKA, seven studies examined a non-selected cohort and one study examined a cohort of patients younger than 75 years old.

The total number of patients in the 13 TKA cohorts was 3261 and the mean age of these patients varied between 49 and 73 years at time of surgery, with ranges from 21 to 96 years. Mean BMI varied from 27 to 34 kg/m² with ranges from 16 to 44 kg/m², but was clearly described in three of the 13 included studies. Only three of the 13 studies provided information concerning possible restricting comorbidities on levels of PA. Five of the 13 studies provided information about the rehabilitation protocols followed.

The total number of patients in the eight UKA cohorts was 662. The mean age at time of surgery of these patients varied between 59 and 72 years, with ranges from 21 to 95 years.

The BMI of the patients was specified in three of eight UKA cohorts and was described as means of the total cohort of 26 and 28 kg/m² (range 20–42) and Pietschmann et al. [46] described a mean BMI in active patients of 28 kg/m² (range 20–56) and in inactive patients of 29 kg/m² (range 19–43). Fisher et al. [51] took into account medical problems restricting PA after UKA and four of eight studies provided information about the rehabilitation protocols. The results of the data extraction are presented in Table 1 for articles concerning data of RTS after TKA and in Table 2 for articles concerning data of RTS after UKA.

Three of the 18 studies, namely Bradbury et al. [50], Huch et al. [44] and Naal et al. [14], were rated as having a low risk of bias, nine were scored as moderate [40, 41, 45‐47, 49, 51, 52, 54] and six as high [42, 43, 48, 55]. It was notable that most studies provided no information about possible confounding factors, as six studies scored ‘high’ and eight studies ‘moderate’. The lowest risk of bias was found for the prognostic factor in which the type of prosthesis was described. No study scored a ‘high’ risk of bias in that domain. A summary of all scored risks of bias per domain is listed in Table 3.

Eight of the 13 studies reported data about the percentages of patients who RTS after TKA. Mean percentages of RTS varied from 36 to 89 %. Huch et al. [44] described two possible percentages of RTS depending on which moment the preoperative sports participation percentage was chosen. They found that 94 % of the patients were active in sports preoperatively ‘during life’, but only 36 % of the patients were still active in sports ‘at time of surgery’. Hence, the rates of RTS after TKA compared with these two different preoperative percentages were 36 and 81 %, respectively. Nine of the 13 TKA studies clearly defined the time period of scoring the preoperative sports participation. Argenson et al. [42] defined the preoperative sports moment as ‘at the time of surgery’ (RTS 86 %), Wylde et al. [53] used ‘3 years before surgery’ (RTS 73 %) and Lefevre et al. [43] and Huch et al. [44] used ‘participation during life’ (RTS 63 and 36 %, respectively) as the preoperative sports moment.

Seven of the nine studies reported the overall percentages of patients who RTS after UKA. Mean percentages of RTS varied from 74 % to more than 100 %, meaning that more patients participated in sports postoperatively than preoperatively. Four of these seven studies clearly described that the time period for the preoperative sports participation level was at the ‘presymptomatic phase’ with described RTS percentages of 93, 95, 98 and 75 % [14, 47, 51, 53].

Regarding specific outcome measures of PA after TKA, UCLA scores were retrieved from three studies. Chang et al. [49] described a mean UCLA score of 4.5/10 (4.5 from a maximum possible score of 10) preoperatively and 4.8/10 postoperatively, and 9 % had a score higher than 6/10. Keeney et al. [55] described pre- and postoperative scores of patients younger than 55 years of 3.4/10 and 4.6 /10, respectively, and of patients between 65 and 75 years of 3.8/10 and 4.9/10. Bock et al. [41] described only a mean postoperative score of 5.9/10 in active patients. The Tegner-Lysholm score was described twice; it was 1.3/10 preoperatively and 3.5/10 postoperatively, with 24 % of scores higher than 5/10 in the study by Diduch et al. [54] and postoperatively a score of 3.9/10 was described in the study by Bock et al. [41]. The Grimby score was described only postoperatively in two studies; twice with scores of 2.8/6 [38, 40].

Regarding specific outcome measures concerning PA after UKA, two studies retrieved UCLA scores. Fisher et al. [51] scored a mean score of 4.2/10 before surgery and a score of 6.5/10 after surgery. Walker et al. [47] measured a mean score of 5.3/10 preoperatively and 6.7/10 postoperatively, with two-thirds of the scores >7. Walker et al. [47] also scored a Tegner of 2.9/10 preoperatively and 3.5/10 postoperatively. A Grimby score was measured in only one article, which showed a score of 3.9/6 postoperatively [38]. These PA scores show that after TKA, patients can regularly return to mild-to-moderate activities and UKA patients can return to moderate-to-high activities.

Eight of the 18 included studies described information about the rehabilitation protocol followed after KA, typically not much more than mentioning that ‘full weight bearing was allowed’ or ‘all patients underwent standardised rehabilitation’ (not otherwise specified). Naal et al. [14] and Lo Presti et al. [48] gave the best descriptions by saying that patients were advised not to RTS before a sufficient muscular recovery of both quadriceps and hamstrings was reached.

Whether confounders were taken into account concerning RTS after KA was scored separately in our data extraction form (Tables 1 and 2). Five of the 18 studies adjusted for confounding: Bradbury et al. (50) found negative influences of restricting co-morbidities and complications, and positive influences of motivation and preoperative sports level on RTS, the latter confirmed by Naal et al. [14].

Age was mentioned as a possible confounder in eight studies, but only in five of these studies was this confounder adequately adjusted for; in four studies age did not have any influence on RTS and only Naal et al. [14] reported a negative influence of older age on RTS after UKA. Chatterji et al. [40], Huch et al. [44], Keeney et al. [55] and Wylde et al. [53] found an influence of sex—men were more able to RTS than women—but Naal et al. [14] did not find any influence of sex. A negative influence of high body weight on RTS was described in four studies. Three studies [44, 47, 53] listed specific patient-reported reasons for restricted sports participation after KA. Discouragement from their surgeons, mainly to high-impact types of sports, was one reason, in addition to pain, functional problems, instability and loss of motivation or loss of confidence. Moreover, the importance of counselling advice from the surgeon was mentioned in six studies, in four of which it was explicitly stated that patients were advised not to resume high-impact sports after KA. Only in the article by Lefevre et al. [43], concerning the judokas, did the influence of this advice seem low because many patients returned to sports despite the surgeon’s recommendations to the contrary. This therefore suggests a positive influence of motivation on RTS.

Patients are able to return to both low- and higher-impact sports after both TKA and UKA, with overall percentages varying from 36 to 89 and from 74 to >100 %, respectively. Participation in sports seems more likely after UKA than TKA, with mean total numbers of sports postoperatively of 1.1–4.6 sports per patient after UKA and 0.2–1.0 after TKA. RTS after TKA for low-impact sports was 94, 64 % for intermediate-impact sports, and 43 % for high-impact sports. For UKA, these numbers are 93, >100 and 35 %, respectively. These findings were confirmed by the PA scores of patients, which are higher after UKA than after TKA, namely return to mild-to-moderate activities after TKA and return to moderate-to-high activities after UKA. Time to RTS took 13 weeks after TKA and 12 weeks after UKA, with 95 and 91 %, respectively, concerning low-impact sports.

A common limitation to all systematic reviews, including ours, is that some papers were overlooked. To overcome this problem, we performed an extensive search with sensitive search criteria and synonyms, and by making use of the expertise of a clinical librarian. Another limitation of this systematic review was that it consists of studies with broad heterogeneity in investigated study populations, defined baseline characteristics, chosen outcome measures and, of course, in research quality. Although systematic reviews with meta-analysis are generally seen as ‘a high quality of evidence’, we believe that given these limitations, our findings are at most of moderate quality. According to the outcome measures, many self-designed sports questionnaires were used. This kind of research is prone to so-called ‘recall bias’, as many rely on the patient’s ability to describe their sporting activities of several years previously. Moreover, different PA outcome measures were described, which were mostly not validated. The UCLA scale was most commonly used. Although the intrinsic disadvantage of the UCLA is that it is a categorical measure, it is a validated scale and until 2009 it seemed to be the most appropriate scale available for assessing PA levels in patients undergoing joint arthroplasty [56].

With respect to confounding factors, it is notable that in only seven of the 18 included studies was there a clear definition of the time of assessment of the preoperative sports level given. Considering the definitions used, such as ‘at time of surgery’, ‘at presymptomatic phase’ or ‘during life’, this has a significant effect on the reported RTS percentage, as Huch et al. [44] have also clearly shown. Other confounding factors that should have been adjusted for in determining percentages of RTS are sex, BMI, restricting comorbidities, complications, and psychosocial factors such as motivation and kinesiophobia of the patients. Conflicting results of a possible negative influence of age on postoperative activity have been mentioned previously, but the influence of age on RTS was also not clear from our included studies. Only a few included studies adequately adjusted for some or all of these confounding factors, resulting in an assessment of moderate or high risk of bias in 15 of 18 studies. In five of 13 included TKA studies and in four of eight included UKA studies, possible influences are stated concerning advice given by the surgeon that should also be taken into account. It is reasonable to assume that negative recommendations from their surgeons concerning high-impact types of sports will negatively influence a patient’s return to (especially) higher-impact sports, even if the patient had had the intention of doing so. Furthermore, the percentage of RTS is dependent on the preoperative sports level and (sports) rehabilitation.

A strength of the present study is that it provides a systematic overview of the literature concerning RTS and time to RTS after KA, including PA-specific outcome measures, while differentiating between TKA and UKA and pooling all extracted data. For this purpose, we selected only articles containing data of both pre- and postoperative sports participation, time to RTS and/or specific PA measurements. Most other reviews included general knee function scores like OKS (Oxford Knee Score) and the WOMAC (The Western Ontario and McMaster Universities Osteoarthritis Index) tool, which are generally accepted Patient Reported Outcome Measures (PROMs) nowadays. However, recently it has been stated that using these instruments has substantial disadvantages for the assessment of knee function with respect to activity and participation [57].

In 1996, Vail and Mallon [36] stated that published information on sports participation after joint arthroplasty is retrospective, limited in scope and primarily anecdotal in origin.

Almost 20 years later, negative advice concerning high-impact activities after joint arthroplasty is still more speculative than evidence-based. Concerning these sports recommendations, the general consensus is that return to low-to-intermediate-impact sports within 3–6 months is possible without any problems, while high-impact sports should be discouraged and high-contact athletic activities should be avoided [17, 20, 58, 59]. In contrast, the article by Lefevre et al. [43] showed that 63 % of former black belt judokas resumed their high impact sport, and Mont et al. [60] conducted a promising study with high-level tennis players, all of whom were also able to resume their sport after TKA. Although long-term effects of high-impact sports on outcomes of TKA need to be determined, both studies proved that return to high-impact sport is actually possible. The discussion includes risks of instability, periprosthetic fractures, bearing surface wear, early aseptic loosening, and subsequently premature revisions after high-impact sports. If one considers this subject from a purely mechanical point of view, it seems apparent that the bearing surface wear rate is directly related to the cycles of use. However, accumulating data suggest that prosthetic wear is not simply a function of time in situ, but rather a function of use [61]. During activities such as hiking or jogging, between 40 and 60 degrees of knee flexion high joint loads of 5–10 times body weight can occur, something that not all knee designs are capable of absorbing, so high polyethylene inlay stress may occur [62]. While some studies indeed found higher radiological wear and potential implant failure in active patients, they did not show an increase in revision rates due to high activities at mid-term [63]. This means that the feared higher risk for survival reductions after TKA in active patients cannot be confirmed. However, length of follow-up is not yet adequate to be able to make definitive conclusions on this matter [64]. On the other hand, recent advances in implant technology, surgical techniques and prosthetic designs and materials, and survival rates of new and improved types of KAs are promising for patients with high demands [65]. Several systematic reviews have concluded that patients and orthopaedic surgeons do not necessarily worry about the same things after joint replacement surgery, that patients should be encouraged to become active after joint replacement, and that further research in this area should be stimulated [16, 66‐70].

While younger and more active patients who undergo joint replacement may have higher expectations regarding activity, the literature suggests that nowadays they actually do not participate in functional levels of sports so often after knee replacement [20]. For example, Kersten et al. described that almost half of TKA patients did not meet health-enhancing PA guidelines and they were less active as a normative group [71]. After performing our review, the question arises: Is it due to a lack of will on the part of the patients that they are not always active after TKA? Or are they also highly influenced by negative advice from their orthopaedic surgeons regarding return to sports, as well as other possible restricting factors? Due to the fact that fulfilment of patient expectations after KA is considered to be a predictive criterion of satisfaction, the value of exploring a patient’s expectations regarding activities after knee replacement has been proven [72].

Historically, participating in sports after joint replacement has been discouraged. Although evidence on this subject is still sparse and of low quality, we can learn from this review that these negative recommendations are still not evidence-based and that actually it is possible to play many different sports after knee replacement surgery. Since postoperative outcomes and return to preoperative sports activity levels are influenced by many factors, individual characteristics, preoperative lifestyle, sport levels, motivation and patient preferences should be taken into account when one considers recommendations for athletic activity after joint replacement [73]. To optimise results, patients who demand higher levels of activity should be carefully selected. Since it seems that more patients can RTS and also to higher-impact types of sports after UKA in comparison with TKA, the choice of type of implant should also be considered. For individuals with limited anteromedial or only lateral compartment concerning types of OA, ‘as limited prosthetic constraint as possible’ and ‘as much retention of a “normal knee feeling” as possible’ are desirable.

Papalia et al. [70] recently found comparable RTS activity rates in patients undergoing TKA and UKA, but they based their conclusion on only one article. Regarding the results of our extensive systematic review, we, like Boyd et al. [74], tend to recommend a UKA over a TKA when indicated for a patient who wishes to remain highly active in a sport.

Based on this review concerning RTS after KA, we strongly recommend using the same language concerning generalising a clear definition of preoperative sports level in future studies. It seems most rational to define the (real and only) preoperative sports level as the ‘presymptomatic phase’ and not the moment ‘at time of surgery’. In other words, preoperative sports level should be based on the phase when the patient was not yet restricted in participating in his or her preferred sports because of osteoarthritic knee complaints.

There is still no real reliable and valid method to analyse PA levels, although the level of activity seems an important prognostic factor, as well as a valuable outcome measure in the assessment of orthopaedic disorders [75]. PROMs have gained importance in both clinical practice and medical research and not only to patients and clinicians, but also to regulators, policy makers and health insurance authorities. But skewing and ceiling effects of currently used PROMs have been described, so using these would not be sufficient for reporting PA outcomes after KA [76]. So-called performance-based outcome measures (PBMs) are defined as assessor-observed measures of tasks classified as ‘activities’ using the International Classification of Functioning, Disability and Health (ICF) model of the World Health Organisation (WHO). PBMs are strongly related to patient self-efficacy in actual performance of function, and have been suggested as possible complementary objective measurement tools next to existing PROMs [77]. In predicting return to work in musculoskeletal diseases, PBMs were shown to strengthen the prognostic value of self-reporting modestly, from 9 to 16 % [78, 79]. The measurement of physical function is complex since it contains multi-dimensional constructs. After performing a systematic review, Dobson et al. [80] recommend further good quality research in investigating the measurement properties of PBMs in people with hip and/or knee OA. Following this conclusion, we would like to recommend investigating the possible added value of PBMs to currently used PROMs in predicting RTS after KA.

A lack of evidence is also apparent with regard to the rehabilitation of highly functioning individuals and those who wish to RTS after knee replacement, but there are some promising results that support a more aggressive rehabilitative approach [81]. Remarkably, hardly any information concerning rehabilitation could be extracted from the studies included in our review, while this seems to be a significant issue. We agree that muscular rehabilitation is important and Healy et al. [67] stated that stretching and strengthening programmes could enhance athletic performance after knee replacement, which could actually prevent injuries and protect joint reconstructions. We recommend performing more research on the (possibly protective) role of a more extensive rehabilitation after KA.

In the absence of consensus from the literature with respect to long-term survival rates especially after performing high-impact sports, there is a need for good quality prospective trials. From our review, it can be concluded that for some patients, some types of high-impact sports are possible after KA.

In the meantime, the ‘intelligent participation’ recommendations of Kuster et al. [37] should be considered. They do not only look at the impact of the sport on the joints, but also take into account prior experiences and the way a patient will perform his or her sport. If activities such as skiing, hiking or tennis were not to be performed as a regular endurance activity but on a recreational basis only, they would be less harmful. Moreover, when, for example, shortcuts and steep descents are avoided during hiking, walking slowly downhill and using ski poles can reduce knee joint loads by 20 %. It would also be acceptable for skilled skiers to ski on flatter slopes, avoiding hard snow conditions and moguls, for 1–2 weeks per year. However, it would seem unwise to start such technically demanding sports activities after knee replacement, due to higher joint loads in unskilled performers and because of the high risk of injuries such as periprosthetic fractures.