- Top of page
- PATIENTS AND METHODS
Although shoulder disorders may produce little discomfort in a minority of individuals, they almost always impair activities of daily living and frequently are associated with substantial pain and disability (1). Most shoulder pain is caused by disorders of periarticular soft tissue, in particular disorders affecting the rotator cuff, such as impingement syndrome and rotator cuff tears (2). In general, considerable uncertainty exists regarding the management and effectiveness of medical and surgical interventions for these conditions (3).
With regard to surgery, outcomes following intervention are best assessed in the long term. There is little evidence on long-term results of shoulder treatment, and the quality of much of the published research has been questioned (3). One major weakness common to most studies of outcomes following shoulder surgery has been the lack of appropriate independent methods of assessing outcome.
Patients can provide reliable and valid judgments of health status and of the benefits of treatment (4), and in the context of surgery, provide evidence that is independent of clinical opinion (5–8). Such judgments incorporate measures of disability that accompany the underlying condition and that often are neglected by clinical measures while being of utmost importance to patients. In addition, in a number of outcomes studies in both medical and surgical contexts, patient-based measures of health-related quality of life have been shown to increase the feasibility of conducting large-scale trials and studies with long-term outcomes of interest (9–11).
There are 2 main types of patient-based measures: generic instruments intended to be applicable to a wide range of health problems, and condition-specific instruments designed to address patients' perceptions in relation to a single condition. Such condition-specific instruments are intended to be maximally sensitive to the outcomes of concern of more narrowly defined patient groups. The latter also tend to be shorter and more sensitive to the specific changes in health produced by a particular intervention (12). Sensitivity to change (also referred to as responsiveness) is considered to be one of the most important measurement properties of a health outcome measure (13).
The purpose of this study was to compare different forms of assessment: a standard clinical assessment and 2 validated patient-based measures—one specific and one generic—with regard to their performance and usefulness in assessing longer-term outcomes following shoulder surgery for impingement syndrome with or without rotator cuff tear or repair. In particular, the aims were to examine the sensitivity to change of 3 different forms of assessment with reference to each other and among subgroups defined on the basis of 4 broad types of surgery received in relation to the presence of cuff tears: full repair, partial repair, cuff tear/no repair, no tear/no repair.
It was hypothesized a priori that the group of patients who had no evidence of a cuff tear would have better postsurgical outcomes than any of the other groups, at least in the short term, if only because their surgery was generally less invasive.
PATIENTS AND METHODS
- Top of page
- PATIENTS AND METHODS
A prospective study included consecutive patients who underwent surgery for shoulder pain with a provisional preoperative diagnosis of subacromial impingement, partial- or full-thickness rotator cuff tear, or any combination of these. All patients had received previous treatment consisting of intraarticular steroid injections, physiotherapy, or both. In many cases, the treatment had been initiated by the patients' general practitioner. Failure of the nonsurgical treatment to alleviate symptoms beyond a short period of time was an indication for surgery. Surgical details were obtained from operative notes by the surgeon using a standard questionnaire, at the time of the 6-month postoperative assessment. Surgical procedures can be summarized as including arthroscopic or open subacromial decompression with or without partial or complete repair of any tendon tears comprising the rotator cuff.
Ninety-three patients were originally recruited in a preadmission assessment clinic at a specialist orthopedic hospital between mid-April 1994 and mid-August 1995. No patients refused to participate. Patients completed a variety of assessments at the preadmission clinic. These were repeated in an outpatient clinic at 6 months and once again at an average of 4 years following surgery. Assessments included the Constant Shoulder Score (14), which was assessed by one of 2 orthopedic surgeons blinded to the results of the previous assessment, and 2 patient-based measures: the Oxford Shoulder Score (OSS) (15) and the Medical Outcomes Study Short Form-36 (SF-36) general health survey (16). A proportion of patients who were unable (or refused) to attend any of the postoperative surgical assessments received their followup assessment by postal questionnaire only (identical to the one completed by those who came to the hospital).
Preoperative and postoperative assessments.
(Assessments are described in greater detail elsewhere &lsqbr;15&rsqbr;). The Constant Shoulder Score (14) contains subjective assessments of pain and activities of daily living that are allocated 15 and 20 points respectively (out of a total of 100). In addition, objective measurements of active range of motion and power are allocated 40 and 25 points, respectively. Low scores denote poor function and greater severity of symptoms.
The OSS (15) is a condition-specific questionnaire that is completed by patients unaided. It was originally generated from interviews with patients and contains 12 items, each of which has 5 response categories. Scores from items are summed to produce a single score with a range from 12 (least difficulties) to 60 (most difficulties). It has been shown to be internally consistent, reliable (reproducible), valid, and sensitive to clinical change (responsive) at 6 months following surgery. Similar to the Constant Shoulder Score, the OSS is inappropriate in conditions of instability (17).
The SF-36 (16) is a 36-item questionnaire that is widely used as a general health status instrument. It provides scores on 8 dimensions: physical functioning, social functioning, role limitations due to physical problems, role limitations due to emotional problems, general mental health, energy, bodily pain, and general health perceptions. In addition, health change is measured by a single item asking patients to compare their current health status with that of “one year ago.” Scores for each dimension range from 0 (poor health) to 100 (good health).
With regard to all the above methods of assessment, on the infrequent occasion that a patient or surgeon missed a questionnaire item(s), the dimension comprising those items (e.g., physical function) was not computed. This explains some of the slight variation in the numbers reported in tables.
In addition to the measures already described, administered at both followup stages, 1 questionnaire item asked patients to appraise the success of their shoulder surgery and offered 4 possible response categories ranging from “very successful” to “very disappointing.” Two items asked patients to judge the degree of change in their problems related to their shoulder and day-to-day life following surgery and offered 5 possible responses: “much better,” “slightly better,” “no change,” “slightly worse,” or “much worse” than before the operation. The 4-year followup questionnaire also added a question regarding the length of time during which the shoulder had been satisfactory since surgery.
Group comparisons have been made using change scores (absolute score at time 1 minus absolute score at time 2). Although the distribution of absolute scores tends to be quite skewed, change scores are particularly useful, as they generally approximate a normal distribution. In addition, effect sizes were calculated for the Constant Shoulder Score, OSS, and SF-36 dimensions to assess the extent of change in patients' clinical condition between the preoperative stage and each of the postsurgical assessments. Effect size is a standardization procedure that allows the extent of change measured by different instruments to be compared directly (18). It is calculated as the difference between the sample's mean presurgical and postsurgical scores, divided by the standard deviation of presurgical scores. Effect sizes of 0.2, 0.5, and 0.8 are typically regarded as indicating small, medium, and large degrees of change, respectively.
Chi-square and analysis of variance (ANOVA) were used to test for statistical significance (α = 0.05).
- Top of page
- PATIENTS AND METHODS
The 93 patients were grouped into diagnostic categories according to whether or not they had a rotator cuff tear based on preoperative ultrasound reports in conjunction with perioperative reports of any cuff tear repair procedures. By this method, 36 patients (39%) were diagnosed as having evidence of a complete or partial rotator cuff tear. In 53 patients (57%), there was evidence to support a diagnosis of no rotator cuff tear. Information was missing or inconclusive on 4 patients. The majority of patients had signs of impingement preoperatively.
At the time of surgery, the median age of the 92-person sample was 58.4 years (mean 57.8, range 22.3–83.7; SD 12.4) and comprised 61 men and 31 women (66% and 34%, respectively). Operative procedures in relation to rotator cuff tears were categorized as follows: complete repair of full- or partial-thickness cuff tear (n = 19; 21%), partial repair of full- or partial-thickness tear (n = 6; 7%), no repair attempted of full- or partial-thickness tear (n = 12; 13%), no cuff tear/no repair (n = 53; 59%).
The second and most recent followup assessment was carried out during 1999 and occurred at a median of 3.9 years following the patients' original shoulder surgery (range 2.2–5.6 years). By this stage, of the original 93 patients 1 patient had died, 6 were lost to followup, 4 refused to participate, 82 completed the patient-based questionnaire, and 61 attended for full clinical assessment.
The proportion of patients who received a clinical examination as part of the 4-year assessment (n = 61) was compared with those who were assessed by questionnaire alone (n = 21) according to 4 diagnostic/operative groupings: whether patients had received a full repair of a cuff tear, a partial repair of a tear, no repair of an identified tear, or no repair/no tear present. No significant differences were found. Fewer of the clinically assessed individuals said that their shoulder had never been satisfactory since the operation (4/61, 7% versus 3/21, 14%) and a larger proportion of the clinically assessed rated the success of their shoulder treatment as either very successful or fairly successful (51/61, 84% versus 16/21, 76%). Nevertheless, none of these differences approached statistical significance. A comparison of preoperative and postoperative OSS scores also revealed no differences between these 2 groups.
Table 1 provides details of the surgical procedures that were performed according to whether or not the patient's rotator cuff was torn, and Table 2 provides additional details regarding the size of cuff tears in relation to whether or not repair procedures were attempted (where details were available on this subgroup). Massive tears (≥5 cm and/or involving 3 tendons) were significantly less likely to be repaired (and certainly not fully) than smaller tears (χ2 = 14.3, 6 degrees of freedom [df]; P < 0.05).
Table 1. Details of surgical procedures performed in relation to whether or not the rotator cuff was torn*
|Evidence of rotator cuff tear (n = 36)||No evidence of cuff tear (n = 53)|
|Arthroscopic subacromial decompression (SAD)||9 (25)||45 (85)|
|Open SAD and complete cuff tear repair||18 (50)||–|
|Open SAD and incomplete cuff tear repair||6 (17)||–|
|Open SAD only||2 (11)||0 (0)|
|Repair of cuff tear only||1 (3)||–|
|Acromioplasty||0 (0)||1 (2)|
|Other shoulder surgery (e.g., manipulation, distension)||3 (8)||10 (19)|
Table 2. Relationship between size of rotator cuff tear and extent of repair*
|Size of tear||Extent of repair|
|Full (n = 16)||Partial (n = 6)||None (n = 11)|
|≤3 cm (small), n = 8||5 (63)||0 (0)||3 (38)|
|>3 to ≤5 cm (medium), n = 6||5 (83)||1 (17)||0 (0)|
|>5 cm and/or 2 tendons (large), n = 3||3 (100)||0 (0)||0 (0)|
|≥5 cm and/or 3 tendons (massive), n = 16||3 (19)||5 (31)||8 (50)|
Preoperative and postoperative health status scores.
Presurgical, 6-month, and 4-year postsurgical health status scores and effect sizes were compared for patients categorized according to the 4 broad types of surgery received in relation to the presence/absence of cuff tears: full repair, partial repair, cuff tear/no repair, no tear/no repair. None of the preoperative scores revealed any significant differences between the 4 groups based on this form of categorization, although patients who received a partial cuff repair—most usually associated with a massive cuff tear—generally had poorer preoperative scores (worse pain/disability) than any of the other 3 groups.
Constant assessment scores were consistently poorer at 4 years than at 6 months: the Constant mean score (SD) at 4 years/6 months for full repair was 67.9 (19.1)/86.7 (17.0); for partial repair it was 56.0 (19.5)/62.2 (31.3); for no repair it was 72.7 (14.0)/83.7 (26.2); for no tear it was 78.8 (18.2)/88.6 (15.6). Patient-based scores appeared remarkably stable when comparing 6-month and 4-year postoperative scores on the same scoring systems and relevant domains: OSS mean score (SD) at 4 years/6 months for full repair was 21.9 (9.4)/22.0 (8.2); for partial repair it was 35.5 (13.6)/35.4 (16.8); for no repair it was 24.4 (10.5)/21.2 (6.2); and for no tear it was 19.7 (9.4)/21.4 (8.3). The SF-36 mean physical activity score (SD) at 4 years/6 months for full repair was 73.3 (23.9)/71.9 (27.8); for partial repair it was 61.3 (37.1)/47.0 (34.9); for no repair it was 50.0 (25.0)/47.7 (28.4); and for no tear it was 53.3 (17.2)/66.8 (27.0). The SF-36 mean pain score (SD) at 4 years/6 months was for full repair 60.8 (31.5)/65.5 (22.5); for partial repair it was 46.3 (30.2)/42.2 (28.8); for no repair it was 54.4 (15.2)/60.0 (34.8); and for no tear it was 65.0 (26.7)/61.0 (25.4). The SF-36 mean social function score (SD) at 4 years/6 months was for full repair 79.7 (27.0)/77.2 (24.1); for partial repair it was 63.0 (38.9)/44.4 (31.4); for no repair it was 83.8 (19.5)/65.6 (42.7); and for no tear it was 83.2 (24.1)/80.4 (27.3).
The 6-month postoperative Constant and OSS scores, and SF-36 physical activity and social function domain scores were consistently and significantly worse for the group of patients who received a partial repair for a rotator cuff tear by comparison to all other groups (where F is the ratio of the between-groups versus within-groups variability in which the null hypothesis ratio should equal 1; ANOVA: Constant F = 3.6, P < 0.05; OSS F = 4.0, P < 0.05; SF-36 physical activity F = 2.9, P < 0.05; SF-36 social function F = 2.8, P < 0.05). This group's scores were also poor in absolute terms, and additional evidence for the relatively poor outcome that this group experienced was obtained with reference to the effect sizes where, for all scores and domains that showed significant differences between the 4 groups, the effect size was either moderate (∼0.5) or small (∼0.2) for the “partial repair” group by comparison with enormous effect sizes (>1) for the other 3 groups. The only very large effect size (of 1.17) for this group occurred at 6 months in relation to the social function domain of the SF-36. However, in this instance, the effect size represented deterioration rather than improvement (the mean score having fallen from 79.6 at the preoperative state to 44.4 at 6 months postsurgery).
At 4 years following surgery, only the Constant and OSS shoulder-specific measures revealed any differences between the 4 groups (mean scores and SDs given above). Once again, the patients who had received a partial repair of a rotator cuff tear had the poorest scores and effect sizes relative to the other groups and in absolute terms. Effect sizes at 4 years representing the magnitude of change from the preoperative score: Constant full repair = 0.06, partial repair = 0.14, no repair = 1.02, no tear = 1.22; OSS full repair = 1.88, partial repair = 0.14, no repair = 1.02, no tear = 1.55.
Success rating and transition items.
The majority of patients (>80%) rated their operation as either “very successful” or “fairly successful” at both 6-month and 4-year followup assessments, and also responded that their shoulder problems and changes in day-to-day life were either “much better” or “slightly better” since having their shoulder surgery (Table 3). There were 7 patients for whom this information was unavailable at 4 years.
Table 3. Patients' ratings of the success of treatment and change in relevant health status*
|Question and response||6 months||4 years|
|How successful would you say your shoulder surgery has been?|| || |
| Very successful||38 (43)||38 (46)|
| Fairly successful||38 (43)||29 (35)|
| Not very successful||7 (8)||10 (12)|
| Very disappointing||6 (7)||5 (6)|
|How have the problems related to your shoulder changed since having the operation?|| || |
| Much better||53 (58)||24 (29)|
| Slightly better||25 (28)||32 (39)|
| No change||6 (7)||14 (17)|
| Slightly worse||4 (4)||8 (10)|
| Much worse||3 (3)||5 (6)|
|How has the shoulder operation changed your day-to-day life?|| || |
| Much better||50 (55)||50 (61)|
| Slightly better||22 (24)||12 (15)|
| No change||14 (15)||13 (16)|
| Slightly worse||3 (3)||4 (5)|
| Much worse||2 (2)||3 (4)|
We first examined the relationship between these responses and the perioperative assessment of the size of patients' cuff tears (defined as in Table 2: none, small, medium, large, massive). There was no correspondence between the size of a tear and patient's appraisal of their outcome at 6 months or the success of the operation at 4 years. However, at 4 years, those who had had the largest cuff tears were less likely than others to say that their shoulder problem or day-to-day life was “much better” or “slightly better” since surgery (shoulder problems much/slightly better: no cuff tear 33 [73%], small 5 [71%], medium 6 [100%], large 2 [67%], massive 5 [33%], χ2 = 12.80, 4 df, P = 0.012; day-to-day life much/slightly better: no cuff tear 37 [82%], small 5 [71%], medium 6 [100%]; large 1 [33%]; massive 7 [47%], χ2 = 12.56, 4 df, P = 0.014).
Table 4 presents a breakdown of the responses to success and transition items obtained at 6 months and 4 years in relation to the 4 surgical groupings previously described. Overall, the poorest outcomes are represented by the small group of patients who underwent a partial repair of their cuff tear (plus subacromial decompression). More than one-half of these patients said that their shoulder problems and day-to-day life were the same or worse at 4 years following surgery. The other 3 groups exhibited variable amounts of change in their responses spanning 4 years. Those who had undergone subacromial decompression with no evidence of a cuff tear represented the only group where the small number who experienced deterioration following surgery had reduced at 4 years rather than increased.
Table 4. Patients' overall judgment of the success of their shoulder surgery and retrospective assessment of change in their condition*
|Question and response||Tear fully repaired||Tear partially repaired||Tear not repaired||No tear/no repair||Statistical significance (χ2)|
|6 months (n = 18)||4 years (n = 17)||6 months (n = 5)||4 years (n = 6)||6 months (n = 12)||4 years (n = 12)||6 months (n = 52)||4 years (n = 46)||6 months† comparison||4 years‡ comparison|
|“How successful do you feel your shoulder operation has been?”|| || || || || || || || || || |
| Very or fairly successful||17 (94)||16 (94)||2 (50)||3 (50)||11 (92)||7 (67)||44 (85)||39 (85)|
| Not very successful or very disappointing||1 (6)||1 (6)||2 (50)||3 (50)||1 (8)||4 (33)||8 (15)||7 (15)||NS||χ2 = 7.8 P = 0.05|
|“How have the problems related to your shoulder changed since you had your operation?”|| || || || || || || || || || |
| Much better||12 (63)||12 (71)||0 (0)||1 (17)||8 (67)||8 (67)||31 (60)||34 (74)|
| Slightly better||5 (26)||3 (18)||4 (80)||1 (17)||2 (17)||1 (8)||14 (27)||8 (17)||NS||χ2 = 15.0 P < 0.05|
| Same or worse||2 (11)||2 (12)||1 (20)||4 (67)||2 (17)||3 (25)||7 (14)||4 (9)|
|“How has your shoulder operation changed your day-to-day life so far?”|| || || || || || || || || || |
| Much better||8 (42)||11 (65)||1 (20)||1 (17)||10 (83)||7 (58)||30 (58)||30 (65)|
| Slightly better||8 (42)||2 (12)||1 (20)||1 (17)||0 (0)||1 (8)||12 (23)||8 (17)||χ2 = 13.2 P < 0.05||NS|
| Same or worse||3 (16)||4 (24)||3 (60)||4 (67)||2 (17)||4 (33)||10 (19)||8 (17)|
- Top of page
- PATIENTS AND METHODS
The purpose of this study was to compare different forms of assessment: a standard clinical assessment and 2 validated patient-based measures (one specific and one generic) with regard to their performance and usefulness in assessing longer-term outcomes following shoulder surgery for impingement syndrome with or without rotator cuff tear or repair.
Results presented here are not intended to influence how surgeons might treat shoulder impingement conditions with or without rotator cuff tears. Such a study would require a much larger sample size with exacting measurement of the amount of rotator cuff damage carried out prospectively. If feasible, randomization of treatment allocation would, of course, enormously increase the confidence attached to any such study findings. It is hoped that those who carry out future larger-scale clinical trials of alternative forms of treatment might benefit from the evidence presented here in choosing the most appropriate outcome measures.
One of the problems in assessing the effectiveness of treatment for shoulder conditions concerns the long-term nature of outcomes of importance. Aside from the additional cost to the health service of providing extra clinic appointments, patients are frequently unable or unwilling to return to the hospital for a followup assessment some years after treatment. In addition increasing numbers of patients move away or are likely to be lost to followup over time. Another problem in assessing the effectiveness of treatment has been the lack of appropriate independent methods of assessing outcome.
Although a number of shoulder assessment questionnaires have been developed in recent years (19–22), it is rare for these instruments to be generated from interviews with patients in addition to their demonstrating necessary measurement properties comprising reliability, validity, and responsiveness to clinical change. In addition, the assessment of responsiveness, where demonstrated at all, has a tendency to be carried out in the short term following an intervention.
All of the health status measures employed here have been evaluated in previous studies for use as outcome measures in the short term. This study addresses specific aspects of the usefulness, particularly in relation to responsiveness, of these measures over a longer period of time. The patient-based measures had a much higher response rate than the clinical measure, due to the ability of these measures to be completed by mail. This is an obvious advantage, and an important one; however, an even more important issue concerns the competence of the patient-based measures to judge outcomes satisfactorily by comparison with a clinical assessment.
The clinical categorization system adopted here is far from perfect, but is nevertheless a useful device for the purpose of comparing the sensitivity of different measures. According to this schema, the majority of patients (57%) in the sample had subacromial impingement with no evidence of a rotator cuff tear. These patients received a combination of surgical techniques by arthroscopic means: subacromial decompression/acromioplasty, plus or minus distension or manipulation. All other patients (further subdivided into 3 groups) had evidence of a complete or partial rotator cuff tear, generally in addition to having signs of impingement, and only one-quarter of these patients received surgery by arthroscopic means. One-third of the patients with a cuff tear had no repair attempted.
It had been anticipated that those without a cuff tear would have better postoperative outcomes. This group had the least severe symptoms at the preoperative stage according to all forms of assessment and did indeed have better absolute scores at the 6-month assessment and at the 4-year assessment. They did not necessarily experience the greatest degree of change in their symptoms, as evidenced by effect sizes. A more striking finding concerned the very small number of patients who had had a cuff tear only partially repaired. This group had markedly poorer scores postoperatively by comparison with all other surgical groups. They also experienced the least amount of change in their symptoms. The effect sizes were particularly helpful in highlighting this finding. The size of the tear was a likely confounder here because the largest tears were most commonly only partially repaired or not repaired at all.
At 4 years following surgery, only the Constant and OSS shoulder-specific measures exhibited sufficient sensitivity to reveal any differences between the 4 defined clinical groups. This finding was not surprising because the SF-36 is a generic measure and, as such, will measure change due to the occurrence of any changes in health (“noise”) affecting any parts of the body (9). Such interference might have been expected after a period of 4 years.
Overall, the patient-based measures exhibited considerable stability between the 2 followup assessments. The Constant clinical assessment resulted in poorer average scores for all patient groups at the 4-year assessment than had been the case at 6 months postsurgery.
More than 80% of patients rated their operation as at least “fairly successful” at both 6-month and 4-year followup assessments. The majority of patients also said that their shoulder problems were better than before surgery. A comparison of patients' responses to these satisfaction and transition items according to their surgical categorization produced a similar picture to impressions obtained using the other forms of assessment—but particularly the patient-based measures. That is, outcomes were fairly stable between 6 months and 4 years with improvement more likely than deterioration after 6 months; furthermore, people undergoing partial repair of a cuff tear experienced the least likelihood of improvement in the short or medium term.
This article underlines the importance of using condition-specific measures when patients are to be assessed beyond the short term. It also contributes additional evidence of construct validity of the OSS to previously published work (15). The OSS achieved a high response rate, in common with the SF-36 generic patient-based measure. In addition, it shared with the clinical assessment a high degree of responsiveness over time.