Does radiolucency really predict loose components in revision shoulder arthroplasty?

From 2010 to 2019, the volume of primary shoulder arthroplasties in Germany increased by approximately 14% each year, and a sevenfold increase in case numbers is expected by 2040 [1]. A massive rise in primary implantations is predicted in the United States as well [2]. This will consequently lead to an increase in shoulder arthroplasty revisions.

The reasons necessitating revision surgery are various, including component loosening, periprosthetic fractures, instability, component wear, or periprosthetic infections [3]. The most common cause, according to registry data from the German Shoulder and Elbow Society, is loosening of the humeral or glenoid component [4]. Determining whether an implant is well integrated or loose not only affects the diagnosis but is also crucial for a well-planned strategy for revision surgery for any complications.

If there is consecutive imaging that shows migration over time, loosening seems obvious. Much more common, however, is the sole evidence of radiolucent lines (RLL) between implant and bone. RLL are described in more than 75% of shoulder arthroplasties after 40 months [5] and over 80% after 10 years [6]. The extent to which these RLL are associated with the need for revision surgery is a subject of ongoing debate. The definitions of loosening based on radiographic findings vary in the literature. Kahn et al. defined a “humerus at risk” if RLL are present in three or more zones [7]. Gonzales et al. defined loosening of the shaft as either migration or RLL of > 2 mm around the whole shaft [8]. Loosening was hereby reported in 6% of the cases; however, revision surgery was only required in 1%. Gazielly et al. reported definite radiographic loosening of the glenoid in 15.5%, but only 2.5% required revision surgery [9]. In line with this, the asymptomatic presence of RLL is not necessarily an indication for revision surgery [7-10].

To date, the diagnostic value of RLL to predict the actual clinical stability of an implant in case of a planned revision surgery, regardless of the underlying diagnosis, remains unclear.

This study was designed to answer the following research questions: Can RLL in preoperative radiographs predict loosening of a shoulder arthroplasty prior to revision surgery? What are further risk factors predicting a loose component in shoulder arthroplasty revision?

RLL are a common radiographic finding after shoulder arthroplasty. Rates up to 80% after 10 years are described [6]. This study, specifically reporting on cases prior to arthroplasty revision, detected RLL in 29% of the glenoid components and 30% of the humeral components. The relatively low rate may likely be explained by a shorter period between implantation and radiographic evaluation in the current study design. Furthermore, this study defiened RLL at a minimum width of 2 mm [8, 12]; this was not consistant in early studies on this topic [13, 14].

Most publications on RLL focused on a consecutive cohort of patients following shoulder arthroplasties. The current study stands out by directly comparing radiographic to intraoperative findings.

To date, it remains unclear how radiographic findings are associated with actual loosening of the implant. Furthermore, the radiographic definition of loosening is inconsistent in the literature. For example, Melis et al. defined loosening as RLL > 2 mm in more than three zones [12], Gonzales et al. as RLL > 2 mm around the whole shaft, and Kahn et al. saw the “humerus at risk” if there were any RLL in more than three zones [7]. However, only a small percentage of the patients with RLL seems to present symptoms or require revision surgery. In the cohort reported by Gonzales et al., only 2% of the patients underwent revision while 6% where classified as loose according to radiographic evaluation [8]. Therefore, the asymptomatic presence of RLL cannot generally be equated with loosening or an indication for revision surgery [7-10].

The current study focused on revision cases. Even though there was a significant correlation between RLL and loosening, the study showed that 13.1% of the humeral components, positive for RLL were still well fixed, which is a clinically relevant amount. Thus, the appearance of RLL correlates with loosening but does not guarantee for loosening of an implant. The predictive value of RLL was dependent on its localization and magnitude. RLL in only one zone were not associated with loosening (p = 0.337); in contrast, RLL in two or more zones were (p < 0.001). Furthermore, an increasing number of zones with RLL was associated with a higher risk for a loose implant (p < 0.001). While RLL in the proximal zones (1 and 7) were more common in general, RLL in the distal zones (3 and 5) showed an even stronger correlation with loosening (Phi 0.536). Thus, loosening of the humeral component is most likely if the RLL are located distally and involve multiple zones.

As with the humeral component, there is no consensus on the classification of radiographic findings and the definition of loosening in glenoid components. Due to the different fixation techniques in anatomic glenoids versus reverse baseplates, it stands to reason to use different systems for both implant types. However, the huge variety of classifications in the current literature reduces the comparability of the results even more [2, 12, 15]. Over all, loosening rates of up to 65% after 15 years are described [16]. In the current study, RLL were present in 29% of the cases. In TSA, RLL were most likely to occur in the superior and inferior, rather than the central areas. This is in line with the typical loosening mechanism described by Franklin et al. in 1988 as the rocking horse phenomenon [17]. A more even distribution of RLL was found in RSA baseplates, with only the cranial baseplate being affected slightly less frequently. Regardless of the type of implant, RLL were strongly associated with loosening. However, 5.5% of the cases with RLL turned out well fixed.

It is under debate if other imaging techniques might improve the non-invasive detection of implant loosening. Gregory et al. reported RLL in 54 of 55 observations from computer tomography (CT) scans (11 shoulders, 5 observers) and a higher intra-observer reliability compared to radiographs [18]. Yian et al. reported on 47 TSA, 47% of the RLL visualized by CT scans were not seen on plain radiographs [19]. However, none of the studies compared the imaging findings to intraoperative proof of loosening. Mallo et al. compared CT scans of painful shoulder arthroplasties to arthroscopic findings and reported a sensitivity of 70% and a specificity of 75% for the detection of loosening by CT evaluation [20]. Newer techniques such as SPECT/CT are only reported in case series, which are too small to provide reasonable evidence [21]. A recent study by Broden et al. investigated CT micromotion analysis; at a 24-month follow-up, RLL were detected in all glenoid components [22]. However, the functional outcome, assessed by Constant- and Oxford-Shoulder Score, improved over time [22]. Thus, those imaging techniques might provide a more reliable and sensitive option to detect radiographic signs of potential loosening in shoulder arthroplasties, but the clinical relevance of this finding remains questionable.

The current study also aimed to assess patient-related factors predictive of a loose component at revision surgery. Bacle et al. described a significant association between RLL and failed previous arthroplasty [23]. In contrast, our data showed no correlation between a previous exchange of implant and loosening (p = 906). The only factors identified to correlate with loosening were a higher age at the time of revision surgery for loosening of the humeral component (p = 0.030) and longevity of the implant for loosening of the glenoid component (p = 0.046).

The results presented in this study need to be interpreted against the background that only patients that underwent revision surgery were included. Another limitation is the variety in revision diagnoses and the inclusion of different implant designs.

The findings of this study confirm that RLL are correlated with implant loosening. However, a clinically relevant amount of implants were intraoperatively stable despite the presence of RLL; therefore, RLL do not guarantee loosening. The current results revealed the importance of a detailed analysis of RLL. Implants were more likely to be loose with the presence of RLL in an increasing number of zones and with humeral RLL being located more distally. Having this in mind, conventional radiographs still play a major role in the diagnostic workup and preoperative preparation in revision shoulder arthroplasty. It may be beneficial to also apply this study design of comparing radiographic results to intraoperative findings to alternate imaging techniques and/or combinations of those in order to investigate their clinical relevance.