Return to sport after shoulder arthroplasty: a systematic review

Patient characteristics

We refer to as ‘athletes’ in this sturdy any individual who practiced sport before surgery, but not necessarily to a professional or competitive level. Out of the 23 studies, 2199 patients were included. Mean age was 68 years (range 18 to 92.6 years, as reported in 22 studies) (Table 2), sex ratio (male/female) was 763/1164 (as reported in 19 studies). The surgery was performed on the dominant/non-dominant shoulder in 913/505 cases (as reported in 16 studies).

Demographic characteristics according to type of arthroplasty are further reported in Table 2. The overall proportion of ‘athletes’ was 38.5%. The proportion of ‘athletes’ was the highest in patients with TSA (56.3% out of 814 patients reported in 12 studies, mean age 64.2 years, range 18 to 88 years); it was 47.5% in patients with HA (out of 457 patients reported in 9 studies, mean age 62.7 years, range 21 to 91 years), and the lowest in patients with RSA with 18.6% of 928 patients (reported in nine studies, mean age 74.2 years, range 47 to 92.6 years). Differences in proportions of ‘athletes’ between groups were significant (P value range < 0.001 to 0.002).

Diagnosis and treatment

Preoperative diagnosis was reported in 20 studies, and included the following: primary osteoarthritis, post-traumatic arthritis, dislocation arthropathy, rheumatoid arthritis, cuff tear arthropathy, massive rotator cuff tear with or without pseudo-paralysis, proximal humerus fracture, fracture nonunion/malunion, avascular necrosis and tumour.

Concerning the length of follow-up, we recorded an overall mean of 4.2 years (range 0.5 to 12.8 years) with the following distribution: 5.7 years for HA (range 0.5 to 12 years), 4.5 years for TSA (range 1 to 12.8 years) and 3.2 years for RSA (range of 1 to 10 years) (Table 3).

Return to sport

The definition of RTS is variable in the literature. In this systematic review, RTS is defined as the ability to go back to sport (and sometimes a new sport), but not necessarily at the same level as before. Many sports were identified, with (Table 4a) or without load (Table 4b) on the shoulder, and with or without contact. For example: swimming, golf, tennis, hiking, bowling, skiing, jogging, cycling, gym (Table 4a and Table 4b).

In the overall studied population, the mean rate of RTS was only 65.4% (range 23.5–97.3%) (Table 3). On the other hand, in the ‘athletes’ population, the mean rate of RTS was significantly higher at 75.5% (P < 0.001; range 40–97.3%). According to arthroplasty type, it was 69.7% (range 65.5–76%) for HA, 65.1% (range 23.5–93%) for RSA, and 63.3% (range 29.5–97.3%) for TSA in the overall population (Table 3). According to arthroplasty type, it was 71.2% (range 65.5–95.8%) for HA, 77.4% (range 57.1–97.3%) for TSA, and 75% (range 40–93%) for RSA in the ‘athletes’ population (Table 3).

The only significant difference between arthroplasty type was found between TSA and HA in the overall population (P = 0.02). ‘Athletes’ with TSA and RSA had a significantly higher rate of RTS than the overall population (P < 0.001 and 0.015, respectively), whereas the difference was not significant for patients with HA.

The overall mean time before RTS after surgery was 7 months (range 1 to 36 months) (reported in 13 studies). The mean time of RTS according to arthroplasty type was 6.2 months (range 1 to 24 months) for HA, 7.9 months (range 1.5 to 36 months) for TSA, and 6.2 months (range 1 to 36 months) for RSA. RTS time in athletes was shorter overall and for every arthroplasty type except for HA (Table 3).

According to the most frequently practiced sports (Table 4a and Table 4b), RTS after HA was 72.9% for swimming, 66.7% for golf and 60.9% for weightlifting. After TSA, RTS was 86.9% for golf, 84.8% for gym and 82.8% for swimming. After RSA, RTS was 78.4% for swimming, 94.2% for weightlifting and 78.7% for cycling.

The mean RTS at the same level of sport after surgery or better for the ‘athletes’ was 90% for TSA. We did not have enough data to analyse HA and RSA (Table 5). The mean RTS at a better level was 44.9% for HA (only the results of two studies included) and 40% for RSA. The mean RTS at the same level is 77% for HA (two studies included), 40% for TSA (two studies included), and 54.9% for RSA. The mean RTS at a worse level of sport was 9.6% for TSA and 10.6% for RSA (Table 5). These results show that most of the ‘athletes’ regained either their previous level or an improved level when they returned to sport.

RTS after HA

Three retrospective studies by Garcia et al, ¹⁹ Bülhoff et al ²³ and Liu et al ²⁰ report a rate of RTS after HA between 66.7 and 67.2% in 29 to 58 ‘athletes’. Liu et al ²⁰ further reported that RTS is safer after HA than after TSA or RSA because of the lower risk of complications (component failure or loosening). However, long-term results show low satisfaction after HA for primary arthritis (25% at 17 years of follow-up). ³⁴

RTS after TSA

Schumann et al, ¹¹ Bülhoff et al, ¹³ Papliodis et al, ¹⁴ Garcia et al ²⁵ and Mannava et al ²⁷ evaluated patients who underwent a TSA with a main diagnosis of primary arthritis (> 60%). The rate of RTS for the ‘athletes’ was > 85% for all studies except for Bülhoff et al with a rate of 57.1%, and Mannava et al with rates of 69.4% for lower extremity sports and 67.2% for upper extremity sports.

In summary, all studies mentioned above show that patients who were able to maintain a sports activity preoperatively had a greater chance of returning to sport postoperatively. This suggests that failure to RTS was not necessarily linked to the surgery itself but to the severity and length of the preoperative disability. Most patients achieved a full RTS within 6 months of the surgery. A few patients even noticed an improvement in their sports level. However, a small proportion of patients also described persistent restrictions in their sports activity even after their operation.

RTS after RSA

The rate of RTS after RSA is reported at 60–78% in the general population. The main diagnosis in all studies was irreparable rotator cuff tear with or without arthropathy (55.2–76%). The study by Fink Barnes et al ¹⁶ including a significant number of revisions (31% of 78 patients) still found a high level RTS of 79.2%, with 57.7% at the same level in sports mainly consisting of golf, tennis, swimming, skiing and hiking. According to diagnosis, RTS was the highest for primary arthritis (100%), and the lowest for fracture (76.9%). According to sport type, it was the highest for gym (81.5%) and the lowest in tennis (25%). ¹⁷

Overall, there only were few studies looking specifically at RTS in younger patients after RSA. RTS seems to be higher in patients aged less than 70 years compared to patients aged 70 years or over. However, the results of these studies also suggest that care should be taken when returning to moderate and high-demand sporting activities following RSA in the younger population, such as tennis and golf, and that low-impact sports should be privileged (for example: swimming, biking, jogging).

Comparative RTS between HA and TSA

Schmidt-Wiethoff et al ⁸ and McCarty et al ⁹ reported equivalent rates of RTS after HA and TSA between 81% and 84%, whereas Garcia et al ¹⁸ found a higher rate after TSA than for HA (97.3% versus 65.5%, respectively). This discrepancy could be due to the different underlying conditions. The latter study only included patients with primary glenohumeral arthritis, whereas the first two also included various conditions such as avascular necrosis, fracture, and fracture sequelae. Therefore, glenoid involvement may impede RTS after HA.

Level of RTS was higher in sports such as fishing (92%), swimming (75–86%) and downhill skiing (81%), cross-country skiing (78%), golf (77–83.3%), tennis (75–100%), and cycling (100%). ^9,18

Comparative RTS between HA and RSA

Only one retrospective study, that by Liu et al, ²⁰ directly comparted RTS between HA and RSA. They reported a significantly higher rate of RTS after RSA at 85.9% versus 66.7% after HA. According to sport, rate of RTS was 69.2% (HA) and 100% (RSA) for gym, 70% (HA) and 71.4% (RSA) for running, and 60% (HA) and 69.7% (RSA) for swimming (Table 2).

Comparative RTS between TSA and RSA

In a study published in 2018 by Flurin, ²⁶ 377 patients (99 ‘athletes’) underwent a shoulder arthroplasty (169 TSA and 208 RSA). RTS was 29.5% for TSA and 23.5% for RSA. The mean time before RTS after surgery was similar in both groups. The main sports engaged in in this study were swimming, cycling, gymnastics, walking, aqua fitness, Nordic walking, gym, and skiing.

Comparative RTS between HA, TSA and RSA

Only one retrospective study by Wang et al ²⁸ directly compared HA, TSA and RSA, in 109 patients (70 ‘athletes’). The overall rate of RTS for all patients was 49.3% for HA, 70.4% for TSA, and 28.6% for RSA; and for ‘athletes’: 76% for HA, 95% for TSA, 40% for RSA.

We identified two meta-analyses comparing RTS after shoulder arthroplasty. Aim et al ²⁹ looked at nine studies in which subgroup analysis showed a lower rate of RTS after RSA: 76.5% versus HA and TSA combined 80.7%. Unfortunately, they did not carry out any subgroup analyses for HA and TSA because the included studies used mixed data. Liu et al ³⁰ analysed 13 studies and found that the patients with a TSA had a statistically significantly higher rate of RTS than those with HA or RSA (92.6% versus 71.1% and 74.9%, respectively). We found a similar trend in our analysis including 21 studies, although there was no significant difference between arthroplasty types.

The lower rate of RTS after RSA may be due to patients being older, the diagnosis of rotator cuff tear arthropathy leading to lower preoperative function, and the altered natural biomechanics inherent to the configuration of RSA. As stated above, Garcia et al ¹⁸ revealed that age over 70 years was a significant predictor of decreased RTS. Aim et al ²⁹ could not carry out subgroup analyses for HA and TSA because the included studies used mixed data about HA and TSA. However, most studies ^18,30,32,33 conclude that TSA leads to a higher level of RTS than HA, although the latter should be privileged in younger patients with high sports activity demands because of the risk of early glenoid component failure.

In our study, we found a mean rate of RTS in ‘athletes’ that was highest after TSA (77.4%) followed by RSA (75%) and lowest after HA (71.2%) (Table 3 and Table 6). However, time before RTS was longer for TSA (7.9 months), than HA (6.2 months) and RSA (6.2 months).

Our findings confirm that sports placing low demand on the shoulder had a better rate of RTS. For example, swimming had a mean rate of RTS of 72.9% for HA, 82.8% for TSA and 78.4% for RSA. Cycling had a mean rate of RTS of 71.6% for HA, 100% for TSA and 78.7% for RSA (Table 4). However, sports with high demand on the shoulder may also have a good rate of RTS, such as tennis (75% for HA and for TSA, but only 38.1% for RSA). The low rate found for RSA in this specific activity could be explained by the fact that it does not restore as much range of motion as TSA, specifically external rotation, ³⁵ which is important for tennis.

The main strength of this systematic review is that it includes a higher number of articles (23) than previously published similar studies, including recent ones with large numbers of patients, but also articles in foreign languages (French, German). This allowed us to compare TSA, HA, and RSA separately, which was previously not achievable. Also, isolating ‘athletes’ from the general population allowed a more realistic evaluation of RTS, as including patients who practiced little if any sports before surgery would be to underestimate the actual RTS after surgery. Moreover, we have separated the rate of RTS according to different sports, which provides guidance for surgeons when advising patients about time for return to a specific sport.

There were also some limitations. The most important is the heterogenicity of the included studies (age, indication for arthroplasty, type of sport and intensity), their limited level of evidence (all studies are level 3 or 4), and relatively limited number of patients (often fewer than 100, sometimes fewer than 10), introducing a risk of bias. We could therefore not conduct significant in-depth subgroup analyses. For instance, we cannot report on the rate of RTS according to the underlying pathology. Furthermore, although we can affirm that RTS after shoulder arthroplasty is high, there were insufficient data reported on the level of RTS and whether it is sustainable over time. Long-term longitudinal studies are necessary to answer this last question.