Introduction

The primary patellar dislocation refers to a traumatic disruption of the previously uninjured medial peripatellar structures [1]. Patellar dislocation commonly occurs to the lateral side, leading to the injury of medial restraints of the patella, particularly the lesions of medial patellofemoral ligament (MPFL) and medial retinaculum [2–6]. It is a common knee disorder that mostly presents in adolescents and physically active people [3]. Acute patellar dislocation accounts for 2–3 % of acute knee injuries and is the second most common cause of traumatic hemarthrosis of the knee [7, 8]. Patellar dislocation is associated with a high rate of functional impairment [9]. In the long term, patellar dislocation might result in knee pain, recurrent instability, decreased level of sporting activity, and patellofemoral osteoarthritis [10]. For primary patellar dislocation, the best treatment is still a subject of controversy. The non-operative treatment has been historically suggested unless there is an obvious osteochondral fracture or a relevant disruption of the medial stabilizers with subluxation of the patella [7, 9, 11, 12]. Such non-operative management strategies included a short-term immobilization period used for patient comfort, and followed by a formal physiotherapy [13]. However, the results of this approach have been unpredictable. A high percentage of recurrence, instability symptom, and functional disability after non-surgical treatment were reported in previous studies [10, 14]. Due to the unsatisfied curative effect and deeper understanding of the nature of primary patellar dislocations, some authors advocated immediate surgical treatment to reduce the risk of redislocation [5, 15–17]. Various surgical modalities have been reported, including surgical repair or reconstruction of the medial patellar stabilizers (medial retinaculum [18], and medial patellofemoral ligament [19, 20]), distal patellar realignment surgery [21], lateral retinacular release [22], and trochleoplasty [23, 24]. The initial success of these procedures in preventing recurrent dislocation of the patella is well accepted. However, whether surgical treatment can improve clinical outcome compared with conservative treatment is still debated. Much of the evidence supports the value of surgical interventions [25–27], while some trials have not found significant differences between clinical outcomes in the two patient groups [28–30]. The latest published meta-analysis failed to confirm any significant difference in outcome between surgical and non-surgical initial treatment of people following primary patellar dislocation [31]. Therefore, it remains unclear which technique has more advantages.

The purpose of this up-to-date meta-analysis is to compare the effects of surgical versus non-surgical treatment for the primary patellar dislocation, and to provide the best evidence currently available.

Materials and methods

Search strategy

We conducted an electronic searching using the databases including Medline, Embase, and the Cochrane Registry of Clinical Trials. All databases were searched from the earliest records to May 2013. The search methodology was based on the search strategy of Cochrane Collaboration. As an example, an Ovid Medline database search was undertaken using the MeSH terms and text words “Patellar Dislocation”[Mesh], “Patella”[Mesh], “Dislocations”[Mesh], “Surgical Procedures, Operative”[Mesh], “Physical Therapy (Specialty)”[Mesh], “Immobilization/”[Mesh], “Braces” [Mesh], patell$, dislocat$, sublux$, instability, medial patellofemoral ligament reconstruction, medial reefing, surg$, operat$, arthroscop$, non-surg$, non-operat$, and etc. Additional relevant studies were identified by searching the references of included studies, review articles contemporaneously. There were no restrictions on language of the publications.

Inclusion and exclusion criteria

The criteria for acceptance of the studies were as follows: (1) Randomized or quasi-randomized controlled trials (RCTs, or qRCTs) and prospective controlled studies are evaluating operative versus non-operative interventions. No restrictions on study language and surgical or non-surgical treatment strategies were made. (2) The target population consisted of individuals who presented with primary (or acute) patellar dislocation. No restrictions on subject gender and age were made. (3) Surgical versus non-surgical interventions. No restrictions on the surgical techniques or non-surgical treatment strategies were made. Concerning the exclusion criteria, we stipulated the following: (1) Retrospective studies, duplicated studies, abstracts, reviews, and studies without raw data we need. (2) People with recurrent dislocation of patella.

Two researchers reviewed the title, abstract of all retrieved studies. Any potentially eligible study to this subject was then reviewed in full text. All the studies were reviewed independently, with disagreements being resolved by discussion.

Data collection and outcome measures

Once the studies met the inclusion criteria, data were extracted by two independent authors using predesigned standardized data extraction form. Any disagreement was resolved in review groups through consensus. For each trial, we collected data on the following characteristics: study type, sample size, surgical or non-surgical treatment strategies, subject gender and age, and duration of follow-up. The primary outcome variable was the frequency of recurrent patellar dislocation. The other outcome measures of this meta-analysis included Kujala score [32], Tegner activity score [33], Hughston visual analog scale knee score (Hughston VAS) [34], incidence of Grade I or more severe patellofemoral joint osteoarthritis according to the Ahlbäck classification [35], pain score on visual analog scale (VAS), percentage of knees with an excellent or good result according to Kujala score, percentage of excellent or good subjective opinion, percentage of regaining preinjury activity level, and frequency of subsequent surgical intervention. Results with an average of 2–10 year follow-up were gathered for meta-analysis. If standard deviation (SD) for continuous variables were not presented, the raw data were collected from the original articles, or corresponding authors were contacted if necessary.

Methodological assessment

The methodological quality of the included trials was evaluated using the Jadad quality scale [36]. This scale evaluates the quality of randomization, blinding, and reasons for withdrawal or dropout. It gives either a score of 1 point for each “yes” or 0 point for each “no”. A bonus point was awarded if the method of randomization or double-blinding was appropriate, while a bonus point was deducted if the methods were inappropriate. Each study is given a score ranging from 0 to 5 points, with higher points indicating higher quality. Each trial was independently assessed by two authors. All disagreements were resolved by discussion or consultation with a third author. If enough studies (at least 10 trials) were identified, funnel plots were to be used to investigate publication biases [37]. The significance of difference was evaluated with the methods described by Begg and Egger [38]. A value of P < 0.10 for publication bias was considered statistically significant.

Statistical analysis

The statistical analysis was conducted using the software Review Manager 5.2 provided by Cochrane Collaboration. The treatment effects were expressed as risk ratios (RR) with 95 % confidence intervals (CI) for dichotomous outcomes and mean differences (MD) with 95 % CI for continuous outcomes. Heterogeneity was tested using the chi-square test with significance set at P < 0.1. I-square test was also used to quantify the effect of heterogeneity with an I 2 of 50 % or over representing substantial heterogeneity. If there was no statistical evidence of heterogeneity, a fixed-effect model was used; otherwise, a random-effects model was adopted. If SD was required to be calculated from raw data, the SPSS 13.0 software was used.

Sensitivity analysis

Where appropriate, sensitivity analyses examining various aspects of trial and review methodology, including study type, allocation concealment, outcome assessor blinding, and the reportage of surgical experience, were conducted to explore the robustness of the evidence.

Results

Study identification

A flow chart of the study selection process is presented in Fig. 1. An initial search identified 1,275 articles from the search protocol: 62 were from the Cochrane Registry of Clinical Trials; 545 were from Medline, and 668 were from Embase. After further evaluation of the titles, text words, and abstracts, 31 potentially relevant studies were selected for full-text examination. Finally, 10 published articles [25, 26, 29, 30, 39–44] (9 independent trials) were determined as appropriate to mach the inclusion criteria. Among the eligible studies, two published articles were deemed as from a single trial [30, 41], the data at the time of the most recent follow-up were extracted for analyzing [41]. If data were solely reported in the previously published articles, this data were extracted for completeness.

Fig. 1
figure 1

Flowchart of selection of eligible studies

Full size image

A funnel plot analysis was unable to be performed because of insufficient studies identified, as both visual examination and statistical analysis of funnel plots have limited power to detect bias if the number of trials is small.

Study characteristics and quality

The main characteristics of included studies are summarized in the Table 1. Among the included articles, 7 studies were reported to be randomized controlled trials, and the other 2 studies were prospective non-randomized controlled trials. The sample size in the single study ranged from 20 to 127. Most studies on this subject had involved mainly adolescents or adult patients, except one study focused on pediatric patients [42]. All participants of the included trials had primary lateral patella dislocation, and each trial had a comparable baseline features in surgical and non-surgical groups. As describe in Table 1, there are some variabilities of the surgical stabilization procedures and non-surgical immobilization and rehabilitation strategies adopted by the authors. Most surgical procedures focused on repair or reconstruction of medial soft tissues (medial patellofemoral ligament, medial retinaculum, and capsule) and lateral retinaculum release. The other technique like Roux-Goldthwait procedure was performed in one study [44]. The non-surgical treatments were different in the immobilizers, knee position, duration of immobilization, and rehabilitation program. The follow-up period of the trials ranged from 2 to 14 years. The quality assessment results for eligible trials by using the Jadad quality scale are shown in Table 1.

Table 1 Characteristics and methodological quality of the included studies
Full size table

Meta-analysis of clinical results

Frequency of recurrent patellar dislocation

Nine eligible studies provided data concerning recurrent patellar dislocation. The test for heterogeneity demonstrated that no significant heterogeneity existed among these studies (P = 0.17, χ 2 = 11.64, I 2 = 31 %), so, a fixed-effect model was performed. A meta-analysis showed that the incidence of recurrent patellar dislocation was significantly higher in the non-surgical treatment group than that in the surgical treatment group (OR 0.56, 95 % CI 0.37–0.85, P = 0.007; Fig. 2). Sensitivity analysis, with 2 non-randomized studies excluded, revealed that there was also no significant heterogeneity among the RCTs (P = 0.14, χ 2 = 9.64, I 2 = 38 %), and the incidence of recurrent patellar dislocation was still significantly lower in the surgical treatment group (OR 0.50, 95 % CI 0.32–0.79, P = 0.003; Fig. 3).

Fig. 2
figure 2

Forest plot of frequency of recurrent patellar dislocation between surgical and non-surgical management for primary patellar dislocations

Full size image
Fig. 3
figure 3

Sensitivity analysis with non-randomized studies excluded for frequency of recurrent patellar dislocation

Full size image

Frequency of subsequent surgical interventions

Further surgery was defined as any surgical procedure due to patellofemoral problems. The main reason for subsequent surgical treatment was patellar redislocation or instability. Five studies, including 3 RCTs and 2 prospective non-randomized studies, provided data on the delayed surgery rate. There was no significantly statistical heterogeneity among all these 5 studies (χ 2 = 2.53, P = 0.64, I 2 = 0 %), and 3 RCTs (χ 2 = 1.91, P = 0.38, I 2 = 0 %). The meta-analysis of all 5 studies and sensitivity analysis (excluding 2 non-randomized studies), both using fixed-effect model, revealed that the two treatment strategies resulted in comparable reoperation rate (meta-analysis of all 5 studies: OR 1.14, 95 % CI 0.68–1.93, P = 0.61; sensitivity analysis: OR 1.13, 95 % CI 0.62–2.06, P = 0.69; Figs. 4, 5).

Fig. 4
figure 4

Forest plot of frequency of subsequent surgical interventions between surgical and non-surgical management for primary patellar dislocations

Full size image
Fig. 5
figure 5

Sensitivity analysis with non-randomized studies excluded for frequency of subsequent surgical interventions

Full size image

Percentage of excellent or good subjective opinion

Three studies provided data on patient subjective opinion. A pooled excellent and good subjective result from a fixed-effect model (χ 2 = 3.09, P = 0.21, I 2 = 35 %) showed that the satisfaction rate was similar between the two arms (OR 0.55, 95 % CI 0.30–1.01, P = 0.05). Due to lack of sufficient studies, we did not carry out a sensitivity analysis with non-randomized studies excluded (Fig. 6).

Fig. 6
figure 6

Forest plot of percentage of excellent or good subjective opinion between surgical and non-surgical management for primary patellar dislocations

Full size image

Subjective assessment of symptoms and functional outcomes

There are 7 RCTs evaluated knee function using Kujala score. The test for heterogeneity demonstrated that significant heterogeneity existed among the studies (χ 2 = 58.04, P < 0.00001, I 2 = 90 %), so, a random-effect model was performed. The pooled data indicated that two treatment strategies resulted in a comparable score (MD: 5.55, 95 % CI −3.51 to 14.62, P = 0.23; Fig. 7).

Fig. 7
figure 7

Forest plot of Kujala score between surgical and non-surgical management for primary patellar dislocations. *Camanho 2009: SD calculated from the raw data provided in the original study [25]. *Sillanpaa 2009: SD obtained from the previous meta-analysis [31]

Full size image

Two studies graded the subjective assessment of symptoms and functional outcomes using Kujala score, with Kujala ≥85 as excellent/good and Kujala ≤84 as fair/poor. Overall, the pooled data of excellent/good Kujala rate indicated no statistical difference between the surgical and non-surgical arms (OR 2.10, 95 % CI 0.19–23.48, P = 0.55). However, this pooled analysis exhibited statistically significant heterogeneity (χ 2 = 7.56, P = 0.006, I 2 = 87 %; Fig. 8).

Fig. 8
figure 8

Forest plot of the rate of excellent/good Kujala score between surgical and non-surgical management for primary patellar dislocations

Full size image

Two included trials reported Hughston VAS score. The test for heterogeneity demonstrated no heterogeneity existed between the studies (χ 2 = 0.22, P = 0.64, I 2 = 0 %). The pooled result showed a statistically lower Hughston VAS score recorded in surgical arm compared with non-surgical arm (MD: −5.74, 95 % CI −9.46 to 2.02, P = 0.003; Fig. 9).

Fig. 9
figure 9

Forest plot of Hughston VAS score between surgical and non-surgical management for primary patellar dislocations

Full size image

Two studies assessed knee pain using a visual analog scale (VAS). Statistically significant heterogeneity existed between the studies (χ 2 = 5.50, P = 0.02, I 2 = 82 %). A meta-analysis using a random-effects model showed that no statistically significant difference was obtained between the two groups (MD: 0.71, 95 % CI −0.37 to 1.78, P = 0.20) (Fig. 10).

Fig. 10
figure 10

Forest plot of VAS between surgical and non-surgical management for primary patellar dislocations. *Sillanpaa 2009: SD obtained from the previous meta-analysis [31]

Full size image

Figures 11 and 12 illustrated the pooled results of Tegner score, percentage of regaining preinjury activity level. There was no heterogeneity detected (Tegner score: χ 2 = 0.78, P = 0.68, I 2 = 0 %; percentage of regaining preinjury activity level: χ 2 = 0.85, P = 0.36, I 2 = 0 %), and no statistically significant difference of pooled results (Tegner score: MD: −0.46, 95 % CI −0.97 to 0.05, P = 0.07; percentage of regaining preinjury activity level: OR 2.24, 95 % CI 0.90–5.57, P = 0.08) between the two treatment strategies in both fields.

Fig. 11
figure 11

Forest plot of Tegner score between surgical and non-surgical management for primary patellar dislocations. *Sillanpaa 2009: SDobtained from the previous meta-analysis [31]

Full size image
Fig. 12
figure 12

Forest plot of percentage of regaining preinjury activity level between surgical and non-surgical management for primary patellar dislocations

Full size image

Radiographic examination

Two studies assessed the severity of patellofemoral joint osteoarthrosis using the Ahlbäck classification from plain radiographs examinations. Overall, there was no substantial heterogeneity observed between the studies (χ 2 = 0.03, P = 0.86, I 2 = 0 %). The pooled result indicated that the incidence of Ahlback grade I or more severe osteoarthrosis was similar between the treatment groups (OR 3.20, 95 % CI 0.32–32.35, P = 0.32) (Fig. 13).

Fig. 13
figure 13

Forest plot of the incidence of Ahlback grade I or more severe osteoarthrosis between surgical and non-surgical management for primary patellar dislocations

Full size image

Discussion

On the subject of comparing surgical treatment strategies with conservative treatment strategies in the management of primary patellofemoral dislocation, two meta-analysis articles [31, 45] have been published. Nevertheless, there are certain limitations in these previous meta-analysis reports. First, studies evaluating both primary and recurrent patellar dislocation episodes were taken together for analysis in one study [45]. It inevitably increases the clinical bias and makes the interpretation of their results difficult. Non-surgical treatment is unsatisfactory for recurrent or habitual patellar dislocation because the normal anatomy is not restored [28, 29]; so, most experts considered surgical strategy as an effective treatment for these patients. Thus, it is not appropriate to mix the primary and recurrent episodes together. Second, the previously published meta-analysis articles provided controversial results and conclusions. The findings of earlier meta-analysis indicated that surgical management was associated with a significantly lower incidence of patellar redislocation but a higher risk of patellofemoral osteoarthritis [45], while the strength of evidence could be weaken due to the limitation of mixed population analyzed together, which has been mentioned above. Later, Hing et al. [31] published another meta-analysis and concluded that there was insufficient evidence to confirm any significant difference in outcome between surgical or non-surgical strategies. Furthermore, the latest search date was October 2010, and the limited number of trials was identified [31]. Since then, several comparative trials for the treatment of primary patellar dislocation have been published and that allowed to update meta-analysis possible.

In this up-to-date meta-analysis, only trials for patients following primary patellar dislocation were included. By taking this measure, we expected to reduce clinical bias to a minimum and focused on the issue of finding a better treatment strategy for patients following primary patellar dislocation. This meta-analysis included 9 trials, some of which were recently published and not included in previous meta-analysis. According to our pooled results, we found a reduced incidence of recurrent dislocation caused by initially surgical management compared to non-surgical management. In addition, no significant heterogeneity was detected among the included studies in this field, although it existed in previous meta-analysis articles [31, 45]. The reason for this change could be that the retrospective studies were excluded, and new high-quality RCTs were added. The results proved that the measures we took to minimize the clinical bias were effective. As to functional outcomes, we found that surgical treatment was associated with a lower Hughston VAS, which corresponded with the previous studies [35, 49]. However, no difference was found in other functional outcomes including further surgery, patient satisfaction rate, VAS, Kujala scores, Tegner score, and rate of regain preinjury activity level. We still found a notable heterogeneity in several fields such as Kujala scores and VAS, although the data in these fields were extracted from RCTs. This could probably be explained by various factors, such as variability in the study design and methodological quality, target populations, intervention strategies, follow-up duration. Subgroup analysis might be a good choice for resolving the problem. However, due to insufficient number of studies identified, we could not split the participant data into subgroups for analysis. This calls for additional large high-quality trials, and should be investigated further. We did not find significant difference in the field of patellofemoral OA according to Ahlback system, which was contrary to the previous meta-analysis studies [45]. This inconsistency could be due to the inclusion of trials dealing with recurrent patellar dislocation, which were excluded in our study.

It is worth noting that various surgical techniques (MPFL repair, MR repair, capsule repair, LR, and Roux-Goldthwait procedure) and non-surgical treatment regimens (ranging from 3 to 6 weeks’ immobilization, different immobilization methods, and different physiotherapy and muscles strengthen program) were undertaken in the included studies as describe in Table 1. Previous studies have reported that different surgical techniques or non-surgical rehabilitation program, anatomical abnormality of the knee, and MPFL injury at different locations may be associated with the outcomes. Ma et al. [46] found that the medial retinaculum plasty was better than medial capsule reefing in improving the subjective effects and decreasing the rate of patellar instability. Xie et al. [47] reported that MPFL reconstruction with polyester suture augmentation results in better outcomes than that without augmentation. Rood et al. [48] conducted a prospective randomized trial comparing taping and cylinder cast immobilization, and found a better Lysholm score was associated with tape bandage immobilization. Armstrong et al. [49] carried out a no immobilization rehabilitation program of patients following first-time patellar dislocation, and found a trend of superior short-term functional outcomes compared to those immobilized with cast. The investigators also found that individuals with trochlear dysplasia can be treated successfully by surgical treatment [24] and MPFL-VMO overlap-region injury might response better to non-surgical strategy [50]. These reports indicated that more investigations should be made for seeking a more appropriate surgical or non-surgical treatment algorithm for specific populations. This issue was also discussed in the previous meta-analysis [45]. The suggestion of that the prognostic factors (such as family history, anatomical presentation, soft tissue lesion type, and rehabilitation regimes), which might be associated with the overall outcomes, should be taken into account in choosing treatment strategy in future studies seems reasonable.

Certainly, there are some limitations in our meta-analysis. Firstly, the number of trials included was not so adequate, which just had 9 separate trials. The sample sizes of the trials were generally small, and the follow-up durations were different between the included studies as described in Table 1. These factors might weaken the strength of our pooled results. Secondly, some unpublished studies and data were not available. We could not investigate publication bias because a funnel plot can not be done due to the limited trials identified in this study. In addition, the quality of some trials, which did not provided adequate randomization and blinding method, was not high enough. These limitations should be taken into account and should be avoided as far as possible when drafting new trials. More large high-quality randomized controlled trials could overcome the limitations currently present in this study.

Conclusions

The available evidence indicates that surgical management was associated with a lower risk of recurrent patellar dislocation, but a lower Hughston VAS than non-surgical management for primary patellar dislocation. More large high-quality trials and further studies are needed to overcome the limitations presented in this study.