Background

Although arthroscopy of upper extremity joints was initially introduced mainly for diagnostic purposes it is being increasingly used for therapeutic interventions [1]. For example, wrist interventions performed through arthroscopy include, among others, excision of wrist ganglia, treatment of acute fractures and of non-unions, ligament repair and reconstructions, repair or debridement of the triangular fibrocartilage complex, ulnar head resection, partial or total removal of carpal bones, and joint fusions [1, 2]. A recent study on musculoskeletal upper extremity ambulatory surgery in the United States estimated that 272,148 rotator cuff repairs, 257,541 shoulder arthroscopies excluding those for cuff repairs, 3686 elbow arthroscopies, and 25,250 wrist arthroscopies were performed in 2006 [3]. Arthroscopic interventions generally require special equipment and substantial surgical training and may thus be associated with higher costs than open procedures [4]. In addition, arthroscopic procedures may be associated with various complications [5]. Arthroscopic interventions may, however, be more cost-effective if their efficacy is superior to that of non-arthroscopic treatments or if they have similar efficacy but provide additional benefit, such as quicker recovery or lower morbidity. There is strong agreement that good-quality randomized controlled trials (RCTs) are the gold standard for assessing treatment efficacy and that they provide higher level of evidence than observational studies [6]. We reviewed the literature for intervention RCTs involving wrist arthroscopy, and for comparison, shoulder arthroscopy, hypothesizing that the quality of wrist and shoulder RCTs are similar.

Methods

We performed a systematic review of the literature for randomized or quasi-randomized clinical trials in which at least one arm was an intervention performed through wrist arthroscopy or shoulder arthroscopy. An experienced researcher searched for articles published up to December 2012 in the databases PubMed and Cochrane Library. The search was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [7]. The search strategy was applied to PubMed and optimized for the Cochrane database (Additional file 1). We included all RCTs written in English, Spanish, or German. We omitted conference abstracts. We checked the references of the initially included articles to identify other potentially relevant studies and subjected them to a similar selection process.

Three researchers reviewed the selected articles (each article reviewed by at least two researchers) and recorded the following data: the country where the study was conducted, the condition for which the interventions were done, the randomization method, the number of randomized participants, the time of randomization, the outcomes measures used, blinding, and description of dropouts and withdrawals. When appropriate we grouped the conditions for which the interventions were done into diagnostic categories. As a measure of RCT quality we used the Jadad scale [8] as modified by Gummesson et al. [9]. The scale considers the randomization method, blinding and description of dropouts/withdrawals, yielding a score from 0 (lowest quality) to 5 (highest) [9]. A study that describes an appropriate randomization method (such as computer-generated sequence or a random-number table) is awarded 2 points while a study that does not report the randomization method or reports an inappropriate method (such as order of presentation or medical record number) is not awarded any points. Similarly a study that reports blinding (single or double) using an appropriate method is awarded 2 points while use of an inappropriate blinding method or absence of blinding does not yield any points. The blinding method was considered appropriate if the article specified whom the blinding involved and, depending on the nature of the interventions, possible additional measures to ensure the blinding (for example, stating that blinding involved an assessor and that the surgical area was covered during patient assessment or that identical incisions were used for the different surgical procedures). Description of any dropouts or withdrawals (or a statement that no dropouts/withdrawals occurred) is awarded 1 point. The grading according to the modified Jadad scale was done by two researchers independently and any disagreements were resolved by discussion until consensus was reached.

The median modified Jadad scores were calculated for the wrist and shoulder RCTs and were then compared with the Mann–Whitney test. A p-value of less than 0.05 was considered to indicate statistical significance.

Results and discussion

Results

The Medline search showed that the first publications in which wrist arthroscopy or shoulder arthroscopy were mentioned appeared in the late 1970s.

Wrist arthroscopy

Of 7 possible RCTs obtained in the search, 3 were excluded because they involved postoperative analgesia, leaving 4 intervention RCTs eligible for inclusion (Figure 1; Additional file 2). The 4 RCTs (Table 1) involved Kienböck’s disease (arthroscopic versus open surgery), dorsal wrist ganglia (arthroscopic versus open excision), volar wrist ganglia (arthroscopic versus open excision), and distal radius fracture (arthroscopically- and fluoroscopically-assisted versus fluoroscopically-assisted reduction, followed by fixation). The number of participants in the 4 studies was 16, 50, 72, and 40, respectively (median 45).

Figure 1
figure 1

RCTs involving wrist arthroscopy or shoulder arthroscopy – inclusion and exclusion flow diagram. Details of the inclusion and exclusion process of the finally selected intervention randomized controlled trials in which at least one arm involved wrist arthroscopy or shoulder arthroscopy; shown in a PRISMA flow diagram. W = number of wrist arthroscopy articles; S = number of shoulder arthroscopy articles.

Full size image
Table 1 Details of the intervention randomized controlled trials in which at least one arm involved wrist arthroscopy or shoulder arthroscopy
Full size table

Shoulder arthroscopy

Of 130 possible RCTs obtained in the search, 80 were excluded: 24 were not intervention RCTs (matched cohort or cross-sectional studies, non-clinical RCTs, RCT protocols), 10 were systematic reviews or meta-analyses, 32 involved anesthesia or postoperative analgesia, 7 involved physiotherapy/postoperative rehabilitation, 6 were subsequent publications of same RCT, and 1 was not intervention through arthroscopy (after review of full-text and contact with the author). Thus, 50 shoulder intervention RCTs were included (Figure 1; Additional file 2). The 50 RCTs (Table 1) involved rotator cuff tears (n = 22), instability (n = 14), impingement (n = 9), and other conditions (n = 5). The interventions compared were different arthroscopic procedures (n = 31), arthroscopic versus open procedures (n = 12), and arthroscopic procedure versus nonoperative treatment (n = 7). The median number of participants was 60 (range 17–150).

Trial quality

Of the 4 wrist studies 2 used inappropriate randomization methods and the remaining 2 stated use of “sealed envelopes” but without reporting how the randomization sequence was generated. None of the studies reported blinding and only 2 provided information about dropouts/withdrawals. In the 50 shoulder RCTs, the randomization method was described and appropriate in 25 (50%), described but inappropriate in 18 (36%) and was not described in 7 (14%). Blinding using an appropriate method was reported in 23 studies (46%), blinding was reported but the method was inappropriate in 5 (10%) and blinding was not reported in 22 studies (44%). Dropouts/withdrawals were described in 41 (82%).

The median modified Jadad score for the wrist arthroscopy intervention RCTs was 0.5 (range 0–1) and for the shoulder arthroscopy intervention RCTs was 3.0 (range 0–5). The quality of the shoulder RCTs was significantly higher than that for the wrist RCTs (p = 0.012).

Discussion

Our study shows that despite the increasing use of wrist arthroscopy in the treatment of various wrist disorders the efficacy of arthroscopically performed interventions has only been studied in 4 quasi-randomized studies. This can be compared to 50 randomized or quasi-randomized studies of significantly higher quality for arthroscopically performed shoulder interventions, yet both procedures were first described in the literature in the late 1970s.

Since their introduction as diagnostic tools, both wrist and shoulder arthroscopy have undergone technical advancement and broader clinical applications. However, they appear to diverge in the extent to which they have been evaluated scientifically. It might be argued that shoulder disorders are more common and therefore it would be easier to conduct randomized trials. However, wrist arthroscopy is being used for several wrist disorders that are relatively common. Besides, multicenter trials can be conducted when a condition is not that common to allow enrollment of an adequate number of patients in a reasonable time. In contrast to wrist arthroscopy, endoscopic carpal tunnel release, an arthroscopic procedure, first described in the literature in the late 1980s, has been evaluated in numerous intervention RCTs, including a number of high quality trials as judged by the Cochrane reviews [10]. Also, our review of shoulder arthroscopy RCTs shows that it is possible to conduct good-quality surgical intervention trials involving arthroscopy.

Arthroscopic interventions are now used for new areas in upper extremity surgery such as thumb carpometacarpal osteoarthritis, a common condition, still without evidence from randomized studies. Because conducting good-quality surgical RCTs, with the many factors involved, is generally more difficult than pharmaceutical trials, proposals have been presented recently to facilitate surgical trials [11, 12]. The lack of high-level evidence, based on good-quality randomized trials, to support the large number of surgical interventions performed through wrist arthroscopy should be a concern not only to health care payers and providers but also to patients.

Like other quality assessment systems, the Jadad scale has its limitations. Although the scale considers the appropriateness of the randomization method, which is fundamental, it does not include concealment. We have however extracted the data concerning concealment for each trial, when such data were reported (Table 1). Further, blinding of patients may not be feasible in surgical interventions. However, we also considered blinding of outcome assessors and this should be feasible in surgical trials. Another limitation is the possible existence of RCTs that the search did not capture. However, we do not believe that the search missed any eligible wrist intervention RCTs.

It is highly unlikely that a study that had used blinding or achieved complete follow-up with no drop-outs or withdrawals would not report these in the published article as important strengths. We considered studies that only mentioned using “sealed envelopes” without specifying how the randomization sequence was generated (2 wrist studies and 11 shoulder studies) as not having reported the randomization method and thus were not awarded any points for randomization. Even if we assume that these studies had used appropriate methods in generating the randomization sequence the results would be similar (median score 1.5 vs 3.0; p = 0.041).

In our search we could not find any previous studies that have assessed the quality of intervention trials involving wrist arthroscopy. With regard to RCTs that involved shoulder arthroscopy, there have been systematic reviews of intervention trials for specific shoulder disorders that included interventions done through arthroscopy. Most of these reviews used different quality scales and therefore could not be compared directly with our study. For example, a systematic review of interventions for anterior shoulder instability assessed the quality of 3 trials with a 12-item scale that included concealment and blinding (each item scored 0, 1 or 2 for a best possible total score of 24 points) giving them a score of 17, 16 and 15, respectively [13]. The modified Jadad score for the same 3 trials in our study was 3, 2 and 0, respectively, which reflects the fact that the modified Jadad scale focuses on the unambiguous reporting of the fundamental issues of randomization, blinding and drop-outs/withdrawals.

In one previous systematic review that used the original Jadad scale in assessing the quality of 54 rotator cuff RCTs published from 2001 to 2011, the mean Jadad score was 3.0 [14]. The authors concluded that most trials were of high quality (66% had a Jadad score >3.0) but because almost two-thirds of the high-quality studies were nonoperative trials they suggested that the rotator cuff literature lacks high quality RCTs that are relevant to surgical clinical practice [14]. In another report based on the “comparative effectiveness of nonoperative and operative treatments for rotator cuff tears” systematic review of literature from 1990 to 2009, the authors concluded that the “RCT literature was of particularly low quality with high risk of bias from the manner in which the studies had been conducted” [15]. Thus, despite our finding that most intervention RCTs involving shoulder arthroscopy were of significantly higher quality than the very few wrist arthroscopy trials that have been performed, there is need for further improved shoulder surgical RCTs. For example, six RCTs (published since 2002) that have assessed the efficacy of knee arthroscopy in the treatment of osteoarthritis [16] are probably of substantially higher quality than most shoulder arthroscopy RCTs.

In a study that estimated the number of upper extremity ambulatory procedures performed in the United States in 2006, including wrist and shoulder arthroscopic interventions, the authors concluded that the resources utilized by these procedures are substantial and suggested that evidence-based clinical indications and outcomes of many of these upper extremity procedures remain poorly defined [3]. For interventions involving wrist arthroscopy, our systematic review shows that there is currently a lack of good evidence supporting the efficacy of these procedures.

Conclusions

This systematic review revealed that the efficacy of arthroscopically performed wrist interventions has been studied in only 4 quasi-randomized studies compared to 50 randomized or quasi-randomized studies of significantly higher quality assessing interventions performed through shoulder arthroscopy. In order to advance evidence-based care of patients with wrist disorders, there is a need for high-quality RCTs designed to assess the efficacy of the procedures currently performed through wrist arthroscopy.