Setting and Location Where Data Were Collected
Our local Research Ethics Committee registered and granted approval for this prospective, randomized, double-blind trial, which was performed in a university-affiliated hospital. Between September 2001 and January 2005, eighty-eight patients who had undergone a primary relocation of a primary shoulder dislocation in our affiliated emergency department were recruited into the study. Our unit provides the only acute musculoskeletal trauma service for the local adult population, and the patients were therefore a consecutive, unselected series. An independent data-monitoring group reviewed interim analyses of the trial results in September 2003 and September 2004. On the first occasion, continued recruitment was recommended, and on the second occasion, complete follow-up was recommended for two years before the results were reported.
Inclusion Criteria
To be eligible for the study, a patient had to meet the following inclusion criteria: (1) radiographic confirmation of a first-time anterior glenohumeral dislocation caused by a substantial external force applied to the shoulder either in a collision, fall, or other forceful injury; (2) no associated fracture visible on conventional radiographs at the time of presentation; (3) no other axial or appendicular musculoskeletal injury; (3) an age between fifteen and thirty-five years inclusive at the time of the primary dislocation; (4) presentation to our shoulder clinic within two weeks of the primary dislocation; (5) no medical contraindications to general anesthesia or evidence of cognitive impairment; (6) informed consent obtained from the patient or his or her guardian (if the patient was less than sixteen years old); (7) a local residence and the ability to attend follow-up evaluations during the two years after the primary dislocation; and (8) detection of an anteroinferior capsulolabral detachment (a Bankart lesion) on the arthroscopic examination of the shoulder. If no Bankart lesion was detected on the initial examination, our protocol stipulated that an arthroscopic lavage alone should be performed, irrespective of the patient's group randomization. However, the patient would continue with trial follow-up and would be analyzed as part of the group to which he or she was initially randomized (according to standard intention-to-treat methodology).
Exclusions
During the forty-month period of the study recruitment (September 2001 to January 2005), 130 patients with an anterior shoulder dislocation who were in the target age-group and resided locally were initially assessed for their eligibility during the period of the study. Forty-two patients were excluded for the following reasons (Fig. 1): (1) Twelve patients were not eligible because the dislocation was not caused by a substantial external force applied to the arm. Four of them had dislocated the shoulder anteriorly through intrinsic contractions during a seizure. An additional eight patients with atraumatic instability had dislocated the shoulder without substantial external force being applied to the arm. These eight patients all had marked hyperlaxity (a score of =4 of a possible 9, according to the system of Beighton and Horan24,25), and subsequent investigation with magnetic resonance arthrography revealed that none had evidence of a capsulolabral detachment. A Beighton score of =4 was not an exclusion criterion in itself, if the patient had sustained the dislocation from an external injury. (2) Eleven patients who had an additional fracture visible on conventional radiography were excluded. Three of them had an associated fracture of the greater tuberosity, and eight had an osseous glenoid rim fracture. (3) One patient was excluded because another axial or appendicular musculoskeletal injury was present. He had a fracture of the ipsilateral wrist. (4) Three patients were ineligible because of late presentation at the shoulder clinic. They were seen more than two weeks after the primary dislocation. (5) Fifteen patients declined to participate in the study, either because they were unwilling to undergo primary surgical intervention (ten patients) or because they had a strong preference for one of the two forms of surgical treatment that were the subject of the study and thus were unwilling to be randomized (five patients).
Sample-Size Estimations
Sample-size estimations (power calculation) were based on the interim results of a previous study in our unit, in which all patients received standard nonoperative treatment for the primary traumatic anterior shoulder dislocation5. That study concluded that thirty patients should be enrolled into each group to detect large treatment effects on redislocation rates and shoulder instability-specific functional scores, with a 75% difference between groups, and with Type-I (alpha) error set at 0.05 and Type-II (beta) error set at 0.1 (90% power). From the results of the previous study, we were aware that there would be substantial loss to follow-up during the two-year follow-up period in this young, migratory population. We therefore increased the recruitment into each group by 50% to compensate for this attrition of follow-up.
Initial Treatment in the Emergency Department
Closed relocation of the shoulder was performed in our affiliated emergency department, with use of either the Milch or Hippocratic technique26, with the patient under conscious sedation. All patients were instructed to wear a shoulder immobilizer sling until their initial attendance at our shoulder clinic, which was held weekly during the trial period.
Randomization
At the initial consultation in the shoulder clinic, the patient was assessed with regard to his or her eligibility for inclusion into the study by one of three of the authors (C.M.R., P.J.J., and T.O.W.). The nature of the study was explained to eligible patients in detail, both verbally and by supplying them with a detailed information sheet. This included details of the best available evidence of the potential benefits of the two treatment options as well as the known adverse effects. It was clearly stated that the treatment allocation of the patient would be concealed until he or she either had recurrent instability, requiring further investigation and treatment, or had reached the end of the two-year follow-up period without the development of instability. Consent for inclusion into the trial was obtained at an additional consultation after the patient had had an opportunity to discuss the study with his or her peers or with an independent surgeon who was not involved with the study.
Once consent was obtained, the patients were randomized to be treated either by an arthroscopic examination and lavage alone or together with a Bankart repair. Age and gender were previously shown to be the only independently predictive risk factors for recurrent instability in a previous study of the nonoperative treatment of a primary dislocation in our unit8. Computerized randomization was therefore performed with use of weighted minimization27,28, on the basis of these factors. This technique is preferred to stratification in studies with relatively small sample sizes, in which it is important that known risk factors for the primary outcome measure are distributed evenly between the two study groups27,28.
Key demographic information regarding the patient and the injury was recorded at a preoperative assessment by one of us (E.W.) (see Appendix). The senior author (C.M.R.) performed a shoulder examination and graded the severity of ligamentous laxity using the scoring system of Beighton and Horan24,25. Hyperlaxity was defined as a Beighton score of =4 of a possible maximum score of 9. Conventional radiographs were also made to confirm a congruous relocation of the shoulder, to ensure that there was no radiographically visible tuberosity or glenoid rim fracture, and to measure the size of the Hill-Sachs defect6,29.
Operative Techniques
All operative procedures were performed in an outpatient unit by the senior author (C.M.R.). With the patient under general anesthesia, he or she was placed in the beach-chair position and the upper extremity was prepared and draped. A careful examination under anesthesia was performed to determine the direction and severity of instability. To ensure observer and subject blinding, both the lavage and stabilization procedures were performed with use of two arthroscopic portals. The posterior viewing portal was created inferomedial to the posterolateral corner of the acromion, and the anterosuperior instrumentation portal was created lateral to the coracoid process, through the rotator interval.
A complete evaluation of the capsular and osseous lesions was performed. All operative procedures were digitally recorded, and we recorded the Baker grade of the Bankart lesion30,31, the presence and severity of any osseous glenoid rim erosion or avulsion, and the details of the surgery performed.
Most patients had a joint hemarthrosis (Table I), but no patient had a substantial tear of the rotator cuff or long head of the biceps on arthroscopic examination. Minor capsular tears were encountered in a minority of patients and scuffing of the articular surface of the anteroinferior quadrant of the glenoid was common. Neither of these lesions was treated in either group. All patients who were randomized had a capsulolabral detachment (a Bankart lesion), but there was considerable variation in its size and severity (Table I), ranging from minor anteroinferior avulsion to massive unstable detachment extending into the anterosuperior and posterosuperior quadrants as a superior labral anteroposterior avulsion (SLAP lesion) in continuity32. In eight patients (four in each treatment group), there was a small osseous avulsion of the anteroinferior aspect of the glenoid (an osseous Bankart lesion), which was not visible on preoperative conventional radiography, and accounted for <10% of the distal diameter of the articular surface in all shoulders. These osseous avulsions were always in continuity with the capsulolabral sleeve detachment. In addition to the labral avulsion from the glenoid, one patient in the arthroscopic Bankart repair group had a small humeral avulsion of the glenohumeral ligament, which was not treated arthroscopically.
Both groups of patients underwent lavage, with use of 3 L of irrigation fluid to evacuate the hemarthrosis and other debris. The patients who were randomized to receive an arthroscopic Bankart repair also underwent repair of the capsulolabral detachment. No attempt was made to perform a concomitant capsular plication or shift. The labral detachment was completed, if necessary, with a Bankart rasp and electrocautery. The anterior aspect of the glenoid neck was then decorticated with use of a motorized shaver and burr to create a cancellous bed to encourage soft-tissue healing. Three to five holes were drilled from the eleven o'clock to the five o'clock position (of the right glenoid), depending on the size of the detachment of the capsulolabral complex. The drill-holes were placed at the margin of the articular surface to allow recreation of the glenoid concavity.
With use of the single anterior portal, a suture passer (Linvatec, Largo, Florida) was used to deliver a PDS suture (polydioxanone; Mitek, Johnson and Johnson, Berkshire, United Kingdom) through the detached capsulolabral complex. A Panalok absorbable anchor (Mitek, Johnson and Johnson) was placed onto the limb of the suture on the glenoid side, and this was then inserted in the most superior drill-hole. The arthroscopic core suture was then tied on the capsulolabral side, to keep the knot away from the articular surface. The same maneuver was then performed to pass the other anchors and sutures, proceeding in a superior-to-inferior direction.
When the anteroinferior Bankart lesion was associated with a superior labral anteroposterior detachment (a SLAP lesion), the latter was repaired first, to anatomically reduce and stabilize the superior pole of the labrum. To maintain the concealment of treatment allocation during follow-up, additional instrumentation for suture management and anchor placement was performed percutaneously, without formal arthroscopic portal incisions. When the detachment was predominantly anteroinferior, three anchors were used, whereas when there was a SLAP lesion in continuity, one or two additional anchors were placed in the superior quadrant of the glenoid. The average number of anchors used was 4.2 (range, three to five anchors).
Small osseous Bankart lesions, which had not been detected on conventional preoperative radiographs, were repaired by incorporating them into the soft-tissue capsulolabral repair. Sutures were placed through the capsulolabral sleeve superior and inferior to the osseous lesion, and the corresponding anchors were inserted into the area superior and inferior to the area of osseous detachment from the glenoid rim. Tying of these sutures produced approximation of the osseous lesion to the area of detachment.
Postoperative Treatment Protocol
All patients were advised to wear a sling with the arm held in internal rotation, neutral flexion, and neutral abduction for six weeks after the primary dislocation. After the arthroscopic procedure, passive circumduction exercises with the arm at the side (pendulum exercises) and elbow range-of-movement exercises were permitted for twenty minutes, three times daily with the sling removed.
Following sling removal at six weeks, patients were referred for a rehabilitation program under the supervision of a physiotherapist. Between six and twelve weeks after the dislocation, patients commenced active-assisted shoulder range-of-motion exercises, avoiding elevation or abduction of >90° or external rotation of >30°. After twelve weeks, patients were allowed unrestricted range of motion. Isometric rotator cuff-strengthening exercises were commenced at six weeks, and the patient progressed to isotonic exercises at twelve weeks. We advised patients to continue the strengthening exercises within the first year after the injury. We allowed patients to return to general fitness training, including running and noncontact sports at twelve weeks, but we advised them to avoid returning to competitive sports until six months after the dislocation.
Treatment of Patients Who Had Recurrent Instability
Trial patients who subsequently experienced a recurrent dislocation, or who had symptoms and physical signs of recurrent subluxation develop, were investigated with magnetic resonance arthrography and computerized tomography to evaluate the residual soft-tissue and osseous abnormalities, respectively.
These patients were all offered a second operative intervention with an additional arthroscopic evaluation of the shoulder, proceeding to a surgical stabilization procedure. The form of surgical stabilization performed was determined by the residual soft-tissue or osseous pathology, irrespective of the patient's initial treatment allocation. If there was an unhealed Bankart lesion with a well-formed capsule and inferior glenohumeral ligament, and no glenoid rim or large humeral head defects, an arthroscopic Bankart repair procedure, with retensioning of the anteroinferior capsule by its plication and/or superior advancement on the glenoid, was performed with use of three, four, or five Panalok suture anchors premounted with PDS sutures. If the Bankart lesion was unhealed and the capsulolabral complex and inferior glenohumeral ligament were attenuated, an open Bankart repair and capsular shift procedure was performed. If the Bankart lesion was healed, an open inferior capsular shift procedure alone was performed. If there was an additional large glenoid rim defect or an engaging humeral head defect, an open Bankart repair was combined with a coracoid transfer (Bristow-Latarjet) procedure. The physiotherapy protocol for the treatment of patients who were treated operatively for recurrent instability was identical to that used after the arthroscopic treatment of the primary dislocation. The patients who declined additional surgical treatment were offered an additional three-month program of rotator cuff-strengthening exercises, but they were retained under follow-up and offered continued access to our clinic if they had more episodes of instability.
Outcome Assessment
We aimed to evaluate all recruited patients in a shoulder injury clinic at six weeks, three months, six months, one year, and two years after the primary dislocation. A summary of the outcome measurements made at these times is presented in a table in the Appendix. All follow-up assessments were made by a single research physiotherapist to reduce observer error. The patient and the physiotherapist were both blinded to the treatment allocation, until the patient either had recurrent instability, requiring further treatment, or had reached the end of the two-year follow-up period.
At each clinical follow-up appointment, patients were directly questioned about whether they had sustained any further dislocations and about any further physiotherapy or medical or surgical treatment they had received. We also specifically questioned the patients about symptoms of recurrent instability. We recorded the amount of time that the patient had been absent from his or her normal work duties (if applicable), the timing of the return to sports, and whether he or she had any other adverse events or complications. An adverse event or complication was defined as any event that required additional operative or medical treatment.
We performed functional assessments at the initial consultation after the primary dislocation and at six months, one year, and two years after the dislocation. Functional assessment was carried out by asking the patient to complete a questionnaire in a booklet format. This was completed by the patient prior to the consultation with the research physiotherapist and was based on the Disabilities of the Arm, Shoulder and Hand (DASH) outcomes data collection instrument. The questionnaire incorporates the Short Form-36 (SF-36) general health assessment measure33,34, the limb-specific DASH score35, and questions related to the treatment expectations and satisfaction of the patient and complications requiring further treatment. We also incorporated the Western Ontario Shoulder Instability Index (WOSI)36 into the questionnaire. This score was designed specifically to assess shoulder instability, and it has been previously shown to provide a more sensitive evaluation of patients with shoulder instability8,36. The research physiotherapist then consulted with the patient to ensure that the questionnaire had been fully completed and to document the active and passive range of shoulder movement with use of a goniometer. We measured flexion and extension in the plane of the scapula and abduction and adduction in the coronal plane. External and internal rotation movements were documented both with the shoulder in neutral abduction and at 90° of abduction. Thereafter, one of the authors (C.M.R.) consulted with the patient to discuss any problems which the patient had had since the previous review. The author then examined the involved shoulder and performed specific provocative tests for shoulder instability (the anterior apprehension test, load-and-shift maneuver, and tests for posterior and inferior instability)37,38. The results of each of these tests were graded as being positive or negative for instability.
Given the young, migratory study group, we made rigorous attempts to retain contact with the patients by obtaining a wide variety of contact details, including those of their peers and family members. When patients failed to return for their prearranged review appointments, we made at least three attempts to reestablish contact using this information. If patients were unwilling or unable to attend clinical follow-up appointments, we invited them to complete the full questionnaire booklet component (Questionnaire 1) of the study either by telephone consultation or by mail. The defaulting patients who were unwilling to comply with this component of the study were asked to complete a more concise questionnaire (Questionnaire 2), which inquired only about symptoms of instability or dislocation and additional treatment that the patient had received for the shoulder (see Appendix).
The direct health-service costs of all treatment were estimated prospectively by measuring patient-specific health-service utilization, which was documented for all trial participants from the date of trial entry up to two years thereafter. Data regarding health-service utilization included all hospitalization episodes (inpatient and outpatient), the duration of the inpatient stay, and all visits to outpatient clinics. These were valued on the costs per day or visit, which were calculated from the hospital cost statistics of the National Health Service. Operating-room costs were calculated on the basis of an analysis of the time in the operating room, the composition of a typical surgical and anesthesia team, and the equipment and consumables used. All costs are reported according to the (Scottish Center) price base of the financial year 2005 to 2006.
The primary outcome measure in this study was the development of recurrent instability within the two-year period after the primary dislocation. This time interval was previously identified as the high-risk period in which most recurrences occur after nonoperative treatment8. Secondary outcome measures included the functional outcome scores, range of movement, patient satisfaction, and the economic evaluation of the cost of treatment.
The patient was considered to have recurrent instability either when recurrent dislocation was radiographically documented or when clinical features of recurrent anterior instability had developed. For the latter diagnosis to be made, the patient had to describe symptoms of "slipping" or "apprehension" during daily activities or sports, when the shoulder was in an abducted and externally rotated position. This had to be accompanied by both a positive anterior apprehension test and a positive load-and-shift test on clinical testing. Since there was an element of subjectivity in the diagnosis of recurrent instability without a radiographically documented dislocation, we analyzed both the differences in the rate of radiographically confirmed dislocations between the two treatment groups and the rate of all recurrent instability.
Statistical Analysis
We compared differences between the treatment groups using the Student t test for continuous outcome data and the chi-square or Fisher exact test for categorical data. We also assessed the magnitude of treatment effects for categorical data by assessing the relative risk and performing a number-needed-to-treat analysis. We used repeated-measures analysis of variance and one-way analysis of variance to assess the significance of time trends within-groups and between-groups in continuous outcome data. We performed multivariate logistic and linear regression analysis to assess the effect of the minimization variables of age and sex and other preoperative intrinsic, injury, and surgical variables (Table I) on the primary and secondary outcome measures. The p value was set at 0.05 for all comparisons, and all tests were two-sided.
Eighty-eight patients were recruited into the study between September 2001 and January 2005 (Fig. 1). Four patients (three who had arthroscopic lavage alone and one who had arthroscopic lavage and Bankart repair) subsequently either became impossible to contact or declined to participate in any form of follow-up, leaving eighty-four patients. At two years, eleven patients (four who had arthroscopic lavage alone and seven who had arthroscopic Bankart repair) either had moved away and were unable to attend clinical follow-up evaluations or had declined to return for evaluation. Of the eleven patients, seven (two who had arthroscopic lavage alone and five who had arthroscopic Bankart repair) completed responses to the full questionnaire (Questionnaire 1) by means of either telephone consultation or by mail, and the remaining four patients (two who had arthroscopic lavage alone and two who had arthroscopic Bankart repair) completed only the abbreviated questionnaire (Questionnaire 2). There were no significant differences between the two treatment groups with regard to the number of patients lost to follow-up (Fig. 1).
The majority of the recruited patients were men who had sustained the primary dislocation in a contact sport (Table I). There were no significant differences in the demographic features or injury patterns of the two treatment groups.
Recurrent Instability (Forty-two Patients in Each Group)
In the group that had arthroscopic Bankart repair, three (7%) of the forty-two patients with instability who had been followed for two years had sustained a further radiographically confirmed dislocation and no other patient had recurrent instability (Table II, Fig. 2). In the group that had arthroscopic lavage alone, sixteen (38%) of the forty-two patients either had an additional radiographically confirmed dislocation (twelve; 29%) or had symptomatic recurrent subluxation (four; 10%). The risk of both radiographically confirmed instability and the overall rate of instability were significantly lower in the group that had arthroscopic Bankart repair (Table II, Fig. 2), with a 76% reduction in the risk of a radiographically proven dislocation and an 82% reduction in the risk of all instability compared with the group that had arthroscopy and lavage alone. None of the demographic features measured in Table I was independently predictive of recurrent instability on multivariate analysis.
The mean time to the development of recurrent instability was 11.3 months (range, three to twenty-three months) in the group that had arthroscopic lavage and Bankart repair and 13.6 months (range, five to twenty-two months) in the group that had arthroscopic lavage only. All of the recurrences were in male patients, and the mean age of the nineteen patients (sixteen who had arthroscopic lavage only and three who had arthroscopic Bankart repair) who had recurrent instability was slightly lower than that of the patients whose shoulders were stable, although the difference was not significant (Table III). Survival analysis, with the development of recurrent instability as the end point, revealed a significant difference (log-rank test, p = 0.001) between the groups (Fig. 2).
The treatment group allocation was independently predictive of recurrent instability on multivariate analysis. The minimization variables of age and gender were not significantly predictive of instability after arthroscopic lavage and Bankart repair, suggesting a uniformity of treatment effect across age and gender. This is demonstrated when the patients in both treatment groups are separated, according to their predicted risk of recurrence, into high-risk (=50%) and low-risk (<50%) subgroups. This so-called banding of risk is based on the minimization variables of age and gender, which were the two factors that independently predicted recurrent instability in a previous study of nonoperatively treated primary dislocations8. Male patients who are twenty-seven years old or younger and female patients who are sixteen years old or younger have a risk of =50% for the development of recurrent instability. The risk for recurrent dislocation and for all instability was similar in the high-risk and low-risk groups following arthroscopic lavage and Bankart repair, whereas the risk of recurrence was higher in the high-risk group than the low-risk group after arthroscopic lavage alone.
Assuming that the patients in this study who were treated by arthroscopic lavage alone derived no benefit from the procedure (i.e., the rate of recurrent instability was the same as that after primary nonoperative treatment), if all patients with a first-time dislocation were treated by an arthroscopic Bankart repair, 4.7 patients would need to be treated by this method, in a number-needed-to-treat analysis, to prevent one patient from sustaining a further radiographically confirmed dislocation and 3.2 patients would need to be treated by this method to prevent one patient from having recurrent instability develop (Table II). This suggests that if a policy of primary arthroscopic Bankart repair for all young patients with a first-time dislocation was adopted, a substantial degree of overtreatment would be required in patients who might not have instability develop, in order to prevent recurrence in the smaller number who would have this complication develop when treated nonoperatively.
Treatment of Patients with Recurrent Instability
Of the nineteen patients who had recurrent instability (sixteen who had arthroscopic lavage alone and three who had arthroscopic Bankart repair), three declined further surgery despite experiencing multiple episodes of instability. Sixteen patients (thirteen who had arthroscopic lavage alone and three who had arthroscopic Bankart repair) underwent an additional operation. In all patients, the Bankart lesion was found to be unhealed, although only one patient had marked glenoid rim erosion. Nine patients had an arthroscopic Bankart repair, six had an open Bankart repair and capsular shift, and one had an open Bankart repair and coracoid transfer. The secondary arthroscopic Bankart repair failed to restore stability in two patients who had both been treated by primary lavage. Subsequently, these patients were both treated with an open Bankart repair and capsular shift, and they had no additional episodes of instability at the time of the latest follow-up, two years after the third surgical procedure.
Functional Outcome Scores (Forty Patients in Both Groups)
There was a significant improvement in the functional outcome scores (which were both expressed as a percentage deficit in relation to the hypothetical perfect score for a normal shoulder) for both groups within the first six months after the dislocation (p = 0.001), but there was no significant improvement thereafter. No significant differences were detected between the two groups with respect to either the WOSI or the DASH scores preoperatively or at six months and twelve months after dislocation (Figs. 3 and 4). At two years, the mean functional scores for both of these outcome measures were better in the group that had arthroscopic lavage and Bankart repair (Figs. 3 and 4). The nineteen patients (sixteen who had arthroscopic lavage alone and three who had arthroscopic Bankart repair) who had recurrent instability develop had significantly poorer functional scores at two years than the remainder of the patients in both groups who had a stable shoulder within the first two years after the primary dislocation (Table III). None of the other demographic features described in Table I was predictive of the functional score on univariate analysis. When the patients in the two groups with stable shoulders within the first two years after the initial dislocation were compared, there was no significant difference in the WOSI or DASH scores at any of the time-periods. Multivariate analysis (linear regression) controlling for age, gender, and trial group allocation showed that the development of recurrent instability was the only factor that was independently predictive of the functional outcome. There was no interaction between age or gender and functional outcome.
The SF-36 scores in the treatment groups were not significantly different in any of the eight major subcategories during the two years after the primary dislocation (see Appendix).
Patient Satisfaction (Forty Patients in Both Groups)
There was no significant difference in the initial treatment expectations between the treatment groups (Table I). At two years after the primary dislocation, and prior to disclosure of the treatment allocation, patients in the group that had arthroscopic lavage and Bankart repair scored significantly higher (p < 0.001) in the component of the questionnaire that assessed whether their expectations had been met (a mean score of 94.1% compared with 74.8% for the patients who had arthroscopic lavage alone). This difference remained significant (p = 0.044) when the patients who had recurrent instability develop were excluded from the analysis (a mean score of 93.7% for those who had arthroscopic lavage and Bankart repair and 82.8% for those who had arthroscopic lavage alone).
At two years and prior to disclosure of their treatment allocation, the patients who underwent arthroscopic lavage and Bankart repair also scored significantly higher in response to questions related to satisfaction with their current level of symptoms (chi-square test, p = 0.028), and whether they would undergo surgical treatment for the shoulder problem if they could go back in time (chi-square test, p < 0.001).
Range of Movement (Thirty-eight Patients Who Had Arthroscopic Lavage Alone and Thirty-five Who Had Arthroscopic Bankart Repair)
There was no significant difference in the ranges of active or passive flexion, extension, or abduction between the groups at any of the three postoperative assessments made during the first two years after the primary dislocation. Similarly, there was no significant difference in the range of active or passive external or internal rotation with the shoulder in neutral flexion and abduction or 90° of abduction.
Return to Work and Sports (Forty Patients in Both Groups)
The initial timing of return to work was largely dependent on the type of work that the patient normally performed. The largest group of patients in the study consisted of those who performed nonmanual work (fifty-five patients, including twenty-nine who had arthroscopic Bankart repair and twenty-six who had arthroscopic lavage alone), and there was no significant difference between the treatment groups with respect to the timing of the return to normal activities. The individual numbers of patients who performed light and heavy manual labor and those who were unemployed were smaller, but with the numbers available there was no difference in the timing of return to work or normal daily activities.
There was no significant difference between the treatment groups with regard to the total number of days absent from work in the first two years after the dislocation (p = 0.3). The patients who had recurrent instability develop regardless of group had a significantly greater mean number of days absent from work than those who had not had this complication develop (40.0 days compared with 25.22 days; p = 0.049).
For the sixty-five patients (thirty-three who had arthroscopic lavage alone and thirty-two who had arthroscopic Bankart repair) who played contact sports prior to the initial dislocation, there was no significant difference between the groups with respect to the number who returned to their sport and the timing of their return to the sport. Five patients (two who had arthroscopic Bankart repair and three who had arthroscopic lavage alone) discontinued playing contact sports after the primary dislocation; an additional thirteen patients (two who had arthroscopic Bankart repair and eleven who had arthroscopic lavage alone) had given up contact sports by the time of the final two-year follow-up. The overall risk of discontinuing contact sports within the first two years after the primary dislocation was therefore significantly higher after arthroscopic lavage alone compared with that after arthroscopic Bankart repair (relative risk, 3.4; 95% confidence interval, 1.3 to 9.2; p = 0.007).
Adverse Events and/or Complications (Forty-two Patients in Each Group)
One patient in each group had erythema and swelling develop over the arthroscopic portal sites within the first two weeks after surgery. Although no bacteriological growth was isolated in either patient, they were both treated empirically with broad-spectrum antibiotics by their primary care physician. In both patients, the erythema resolved without further treatment. There were no deep infections or new postoperative nerve palsies.
Three patients (two who had arthroscopic Bankart repair and one who had arthroscopic lavage alone) had adhesive capsulitis develop within the first six months after the primary dislocation. This diagnosis was based on the presence of residual shoulder pain and global shoulder stiffness, with severe loss of external rotation. One patient in each group had relatively mild restriction, which responded rapidly to physiotherapy. The remaining patient, who had undergone an arthroscopic Bankart repair, had more refractory stiffness, which did not resolve after three months of intensive physiotherapy. This patient was successfully treated while under anesthesia by a manipulation of the shoulder combined with an arthroscopic release of the rotator interval and a circumferential capsular release.
Economic Assessment (Forty-two Patients in Each Group)
The initial cost of arthroscopic Bankart repair was greater than that of arthroscopic lavage because of the extra instruments and suture anchors required for the procedure. However, the subsequent cost of treatment in the group that had arthroscopic Bankart repair was lower in the first two years after the dislocation (Table IV). This was due to the reduced requirement for treatment of recurrent dislocations, additional operative stabilization procedures, and postoperative physiotherapy sessions. The reduced costs of treatment during the first two years after the dislocation offset the initial higher expenditure, and the overall cost of treatment was therefore significantly lower in the group that had arthroscopic lavage and Bankart repair (p = 0.012; Table IV).
This study demonstrates that patients who are thirty-five years of age or younger and are treated by an anatomic arthroscopic repair of the anteroinferior aspect of the labrum after a first-time shoulder dislocation have a significantly reduced risk of recurrent instability compared with those who have an arthroscopic examination and lavage alone. They also have significantly better shoulder functional scores, have higher levels of satisfaction with their treatment, and are more likely to be playing contact sports at two years after the dislocation. Although initially cheaper, arthroscopic examination and lavage has higher long-term costs because of the costs of treatment associated with the higher rate of recurrent instability. The findings of the study, therefore, support the rejection of our original null hypothesis, and they confirm that an anatomic repair of the Bankart lesion should be performed if a primary arthroscopic intervention is used to treat a patient with a first-time anterior dislocation of the shoulder.
Multivariate analysis revealed that the development of recurrent instability within the first two years after the dislocation was the sole independent predictor of the functional outcome. It is therefore likely that the better overall functional outcome in the patients treated by an arthroscopic Bankart repair was due to the lower rate of recurrent instability, and the level of function in patients with stable shoulders at two years was similar, irrespective of their initial treatment. Since the therapeutic effect from the arthroscopic Bankart repair is independent of age and gender, younger male patients, who have the greatest preoperative risk of later recurrent instability after nonoperative treatment8, have the greatest risk reduction from an arthroscopic repair.
The internal validity of the study was high, since a previous prospective cohort study of nonoperative treatment of patients with a first-time dislocation was used to design the study protocol8, and the two trial groups had very similar demographic characteristics. We achieved good levels of follow-up, despite the young and relatively mobile study group, and our study is the first to incorporate an economic evaluation of direct treatment costs. Previous clinical studies have described a clinically important reduction in the rate of instability when primary arthroscopic intervention was compared with primary nonoperative treatment10-16,39,40 (see Appendix). The unblinded or single-blind design of those studies produces a greater risk of error because of observer or subject bias, and interpretation is also limited by the inability of the studies to discriminate the therapeutic effects of the Bankart repair from those of the arthroscopic examination and lavage of the shoulder joint. The potential for bias in our study was reduced considerably by adopting a double-blind design, in which both groups had identical surgical incisions and postoperative treatment regimens. Since both groups underwent an arthroscopic examination and lavage, we were able to control for the independent therapeutic effects of these interventions. Therefore, this study is the first, to our knowledge, to quantify the therapeutic benefits of the Bankart repair as distinct from those of the arthroscopic examination and lavage alone.
The rate of recurrent instability in the patients who received an arthroscopic Bankart repair was lower than has been previously reported in case series and comparative studies of this technique10-16,39,40. Arthroscopic repair has advanced considerably over the last ten years, with improved instrumentation, implants, fluid delivery systems, surgical techniques, and aftercare. This may explain the improved results that we achieved in comparison with those in the earlier reports. In addition, many previous studies were confined to athletes and military cadets12-15,39, who may have a higher risk of recurrence because of their higher levels of physical activity and functional expectations. Since our patients were recruited from a single center and were unselected, they were more likely to be representative of the general population of young adults with this injury. It is also notable that the outcomes after primary arthroscopic repair in our study are similar to those reported in other studies in which arthroscopic stabilization was used to treat recurrent instability23,41. Therefore, it seems unlikely that performing arthroscopic stabilization as a primary intervention has a clear advantage over the use of this technique as a delayed procedure for patients with recurrent instability after the primary dislocation.
During our study, some patients with a primary dislocation were probably ineligible for recruitment because they received a so-called first-aid relocation of the shoulder and presented late with recurrent instability. There may have been other patients who continued to avoid medical contact despite the development of recurrent instability. The outcome in these patients might have been different if they had been offered a primary stabilization after the primary dislocation. Our single-surgeon design created a greater standardization of the operative technique than would have otherwise been possible, although similar results may not be achieved by other surgeons using different operative techniques.
The double-blind study design had some disadvantages; most notably, it prevented us from incorporating a control group of patients managed nonoperatively. We were therefore unable to compare directly the therapeutic outcome of our primary arthroscopic interventions with the currently accepted standard of treatment for a primary dislocation. The patients who were randomized to receive an arthroscopic examination and lavage cannot be regarded as a nontreatment or sham incision group, since lavage has previously been shown to increase the rate of resolution of the joint effusion, which may facilitate healing of the labral detachment21. The operative procedure may also encourage the patient to reduce his or her level of physical activity and to comply with postoperative rehabilitation protocols9,10,21,42. The magnitude of the therapeutic effect of the arthroscopic lavage is controversial: a quasirandomized study in one center previously found a reduced rate of recurrent instability after the use of arthroscopic lavage9,10, whereas subsequent uncontrolled case series did not replicate these findings5,31,43. In the present study, the rate of recurrent instability after arthroscopic examination and lavage was lower than has been previously reported for patients treated nonoperatively, even in the younger patients who had a higher predicted risk of instability (Table II). Although there is evidence of a therapeutic benefit from the lavage, there is little to support the use of lavage alone, given the much greater treatment effect from an arthroscopic Bankart repair.
No significant difference was detected between the groups with respect to the postoperative range of shoulder motion, but three patients had development of signs consistent with adhesive capsulitis, which resolved with further treatment. It is unclear whether this was a complication of the initial shoulder dislocation or of the arthroscopic intervention.
The overall cost of treatment after arthroscopic repair was lower than that after a lavage, and the increase was mainly due to the greater secondary costs of treating recurrent instability in the latter group. The economic benefit from the arthroscopic repair over lavage is likely to be an underestimation of the total financial benefit, since we measured only direct hospital costs. The additional community costs of absence from work, loss of earnings, transportation costs for follow-up evaluations, copayment for insurance, and primary-care consultations are difficult to quantify because of the different occupations and earning potential of the recruited patients, but it seems likely that there are considerable additional "hidden" community cost savings from the arthroscopic repair.
As most patients who sustain a first-time dislocation are young athletes who play contact sports, the prevention of subsequent instability is advantageous in reducing their requirement for later surgery, which would necessitate an additional period of enforced absence from work and chosen sport. However, the benefits of primary stabilization must be considered together with its drawbacks: it is initially costly, involves overtreatment of individuals who might not have instability develop (as revealed by our number-needed-to-treat analysis), confers no functional benefit in those who do not have instability develop, has a measurable failure rate, and potentially exposes patients to additional complications. We therefore believe that our results do not provide sufficient evidence to support the routine use of prophylactic repair for all primary dislocations.
Our current practice is to counsel all young patients with a first-time dislocation who are referred to our service. We provide these patients with an estimate of their risk of instability from nonoperative treatment8 and from a primary Bankart repair. We also discuss the risk of complications from arthroscopic surgery. We find that primary stabilization is chosen most often by younger (fifteen to twenty-five-year-old) athletes who play contact sports, and, when possible, they are offered surgery within two weeks after the primary dislocation.
Other adjuvant primary treatments may reduce the risk of subsequent shoulder instability. Preliminary reports have suggested that bracing in external rotation after dislocation may reduce the subsequent risk of recurrent instability44,45, although the magnitude of benefit appears to be smaller than that after arthroscopic surgery. If independent evaluations of this technique confirm these preliminary results, additional trials will be required to compare its efficacy against primary arthroscopic repair. Despite the two-year follow-up interval in this study, we think that evaluation of the longer-term prognosis after primary arthroscopic Bankart repair is required to ensure that the development of recurrent instability has not been delayed by our adjuvant primary treatment and to assess whether the risk of longer-term complications, such as posttraumatic glenohumeral osteoarthrosis, is not increased by the procedure.