LATERAL EPICONDYLITIS, or tennis elbow, is a softtissue

308 Factors Associated With Prognosis of Lateral Epicondylitis After 8 Weeks of Physical Therapy Esther J. Waugh, BScPT, MSc, Susan B. Jaglal, PhD, Aileen M. Davis, PhD, George Tomlinson, PhD, Molly C. Verrier, MHSc ABSTRACT. Waugh EJ, Jaglal SB, Davis AM, Tomlinson G, Verrier MC. Factors associated with prognosis of lateral epicondylitis after 8 weeks of physical therapy. Arch Phys Med Rehabil 2004;85:308-18. Objective: To identify key factors associated with outcomes of patients who underwent 8 weeks of physical therapy (PT) for lateral epicondylitis. Design: Multicenter prospective design with inception cohort of lateral epicondylitis patients commencing PT. Baseline clinical examinations were conducted by 1 physical therapist; self-report outcome measures were completed at baseline and 8 weeks later. Setting: Nine private sports medicine clinics and 2 hospital outpatient departments in Ontario, Canada. Participants: Eighty-three patients with unilateral lateral epicondylitis identified by the treating physical therapists. Interventions: Not applicable. Main Outcome Measures: The final scores of the Disability of the Arm, Shoulder and Hand (DASH) questionnaire and a vertical pain visual analog scale (VAS) were used as the dependent variables. Results: The final prognostic model for the 8-week DASH scores included the baseline score (95% confidence interval [CI], 0.34 0.66), sex (female) (95% CI, 3.3 14.5), and selfreported nerve symptoms (95% CI, 0.8 13.8). The model for the 8-week VAS scores included the baseline score (95% CI, 0.01 0.37), sex (female) (95% CI, 0.4 18.2), and self-reported nerve symptoms (95% CI, 4.7 25.5). A subanalysis indicated that women were more likely than men to have work-related onsets, repetitive keyboarding jobs, and cervical joint signs. Among women, these factors were associated with higher final DASH and VAS scores. Conclusions: Women and patients who report nerve symptoms are more likely to experience a poorer short-term outcome after PT management of lateral epicondylitis. Work-related onsets, repetitive keyboarding jobs, and cervical joint signs have a prognostic influence on women. Key Words: Epicondylitis, lateral humeral; Physical therapy; Prognosis; Rehabilitation. From the Department of Physical Therapy (Waugh, Jaglal, Davis, Verrier); Graduate Department of Rehabilitation Science (Jaglal, Davis, Verrier); Department of Health Policy, Management and Evaluation (Davis); Department of Medicine, University of Toronto (Tomlinson); and Toronto Rehabilitation Institute (Davis), Toronto, ON, Canada. Presented in part at the Canadian Physiotherapy Association s 14th Annual National Orthopaedic Symposium, September 2002, Saskatoon, SK, Canada. No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the author(s) or upon any organization with which the author(s) is/are associated. Reprint requests Esther J. Waugh, BScPT, MSc, Women s College Ambulatory Care Centre, Osteoporosis Research Program, 76 Grenville St, 10th Fl E, Rm 1005, Toronto, ON M5S 1B2, Canada, e-mail: e.waugh@utoronto.ca. 0003-9993/04/8502-8046$30.00/0 doi:10.1016/s0003-9993(03)00480-5 2004 by the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation LATERAL EPICONDYLITIS, or tennis elbow, is a softtissue lesion of the tendinous origin of the wrist extensor muscles that results in lateral elbow pain. It is aggravated by wrist and hand movements and can severely restrict job performance, activities of daily living, and leisure pursuits. Allander 1 reported an annual incidence of 1% to 3% in a population of 15,000 subjects. Although 50% of all recreational tennis players can expect to experience this condition at some point in their playing lifetime, tennis players account for only 5% of all lateral epicondylitis patients seen in clinical practice. 2-6 Most cases are associated with work-related activities 7,8 or have no precipitating factors. 9 Lateral epicondylitis occurs primarily in people between the ages of 40 and 60 years, is equally distributed between men and women, and typically affects the dominant arm. 10-14 Despite being among the most common upper-extremity musculoskeletal syndromes, 10,15 lateral epicondylitis is a therapeutic challenge. Recently, studies 16-19 have begun to consistently demonstrate the effectiveness of corticosteroid injections in providing short-term pain relief, albeit with high recurrence rates. However, there is, as yet, no scientific validation for other conservative therapeutic interventions. Labelle et al reviewed 185 articles and stated that due to poor quality [of the clinical trials] and contradictory results it must be concluded that there is not enough scientific evidence to favour any particular treatment for acute lateral epicondylitis. 20(p650) Hudak et al 21 have argued that a lack of knowledge regarding prognostic factors may be contributing to the inconsistent conclusions stemming from these trials. They suggested that unless treatment groups are similar or stratified on key prognostic variables, different results may be attributable more to differences in prognostic variables than to the treatment interventions. 21(p590) This observation may be particularly relevant given the small sample sizes of the majority of the lateral epicondylitis clinical trials. Indeed, results from clinical trials and observations from clinical practice indicate that response to treatment is inconsistent and that recovery times vary widely among individuals. 21,22 The reasons for discrepancies in treatment response and recovery are not clearly understood. Accordingly, it is important that key prognostic factors be identified to explain these discrepancies. Knowing these factors will aid in the comparison of groups in randomized controlled trials (RCTs) to ensure randomization was successful, in the control of confounding factors in future observational studies, and in the prediction of patient outcomes in clinical practice. To identify factors reported to be associated with outcome of conservative treatment, we reviewed the literature on lateral epicondylitis. To date, only 1 prognostic study 23 has been published. It is a retrospective study that reported that low pretreatment pain scores, prolonged duration of symptoms, and being female were associated with a poor outcome. Outcome was measured by having patients report whether they had

PROGNOSTIC FACTORS FOR LATERAL EPICONDYLITIS, Waugh 309 experienced complete recovery, almost complete recovery, or no recovery after their treatment program. Further information regarding prognostic factors was obtained from data collected as part of clinical trials 24 and from theories presented in descriptive articles. This information provided initial evidence for the possible association between several variables and recovery after conservative treatment. These variables included duration of symptoms longer than 3 months, 18 workrelated cases versus sports-related cases, 25 site of lesion (origin vs tendon body of extensor carpi radialis brevis), 26 history of previous episodes, 26 treatment 18 or cortisone injections, 25 concurrent cervical symptoms, 17,19,27-29 articular elbow signs, 30 abnormal neurodynamics of the radial nerve, 31,32 and supinator muscle involvement. 33 Nine factors that appeared to have the strongest association with recovery of lateral epicondylitis and had not been previously adequately investigated were selected for investigation in the present study: age, sex, duration of symptoms, history of previous episodes, mechanism of onset (work vs sport), site of lesion (tenoperiosteal junction vs tendon body of extensor carpi radialis brevis), elbow articular dysfunction, cervical dysfunction, and abnormal upper-limb neurodynamics. The specific objectives of the present study were to provide descriptive information regarding demographic, injury, and physical characteristics of a sample of lateral epicondylitis patients and to identify prognostic factors associated with clinically relevant outcomes after 8 weeks of physical therapy (PT). METHODS Study Design and Participants This was a multicenter prospective study with an inception cohort of lateral epicondylitis patients commencing PT treatment. Eighty-six people (49 women, 37 men), who were recruited from 9 urban sports medicine and orthopedic clinics and 2 hospital outpatient departments in Toronto, ON, Canada, over a 14-month period, volunteered to participate. Participants were considered to have a diagnosis of lateral epicondylitis and were eligible for inclusion in the study if they had tenderness on or near the lateral epicondyle and if pain was present on at least 2 of 3 pain provocation tests: gripping, resisted wrist extension performed with the elbow in an extended position, and resisted middle finger extension performed with the elbow in an extended position. These criteria are described in standard textbooks 34-36 and were used specifically in Haker and Lundeberg s numerous clinical trials. 37 Participants were excluded if they had a history of elbow fracture or surgery, bilateral symptoms, or concurrent upper quadrant pathology or symptoms not directly related to their lateral epicondylitis symptoms. Eligible patients were identified by the treating physical therapists during the initial PT assessment at the participating clinics. Formal written consent was obtained from each participant. The study was approved by the University of Toronto Human Subjects Review Board. Procedures and Outcome Measures After the principal investigator, a registered physical therapist, performed a baseline clinical assessment, the patient completed 2 self-report outcome measures, the Disability of the Arm, Shoulder and Hand (DASH) questionnaire 38 and a pain visual analog scale 39,40 (VAS). This assessment was conducted at the commencement of PT management for lateral epicondylitis before participants had attended more than 3 sessions. Follow-up questionnaires were completed at 8 weeks after the baseline assessment. The 8-week time frame was chosen because it represents a typical time period for PT management of a soft-tissue injury and would, therefore, be representative of the clinical course of recovery for lateral epicondylitis. 34,36 Outcome measures. The 8-week DASH and pain VAS scores were used as the dependent variables in the development of the prognostic models for lateral epicondylitis. The DASH questionnaire was developed to measure disability in persons with musculoskeletal disorders of the upper limb for both descriptive and evaluative purposes. 38,41 We chose this outcome measure instead of a lateral epicondylitis-specific measure because it has been more extensively evaluated and has a smaller standard error of measurement, and its validity is comparable to the joint-specific measures. 42,43 It consists of a 30-item questionnaire with 5 response options for each item. It is scored out of 100, with a higher score indicating a greater level of disability. The DASH has been tested in various populations and has consistently shown good to excellent reliability, validity, and responsiveness. 38,44-46 A Cronbach of.96 has been reported for internal consistency. 38 An intraclass correlation coefficient (ICC) of.93 was reported for test-retest reliability among patients with elbow pathology, 42 and an ICC of.96 was reported in shoulder and wrist and hand patients. 38 The DASH also correlates with other measures of elbow pain, function, and disability, with Pearson product-moment correlation coefficients ranging from.79 to.90. 42,47 Responsiveness of the DASH to both self-rated and observer-expected change in patients receiving care for shoulder and wrist pathologies, as measured by standardized response means, is comparable to or better than other upper-extremity joint-specific measures. 38 The pain VAS is widely used in both clinical practice and research to evaluate a person s subjective experience of pain severity. It has well-established reliability and validity in many patient populations and is considered the most sensitive of all pain rating scales. 39,44,45,48-50 The vertical pain VAS used in the present study (a 100-mm vertical line anchored with words no pain at the bottom and pain as bad as it can be at the top) has been specifically evaluated in the lateral epicondylitis population and was reported to have an ICC of.89 for test-retest reliability on measurements performed 4 days apart, moderate Pearson correlations with a pain-free function scale (r.68) and pain-free grip strength (r.47), and an ability to detect change in pain when it had occurred as described by both the patient and an external assessor. 40,46 Baseline clinical assessment. The purpose of the clinical assessment was to obtain baseline information regarding the demographic, injury, and physical characteristics of the sample of patients with lateral epicondylitis at the commencement of PT. Table 1 summarizes the variables collected for descriptive purposes and those collected as potential prognostic variables. The categorization of the prognostic variables for the purposes of analysis is also indicated. Information regarding age, sex, symptom description, duration of symptoms, mechanism of onset, history of recurrences, history of previous treatment, hand dominance, job duties, and sport or other leisure activities was obtained from an interview conducted by the principal investigator who asked a series of standard questions before conducting the physical examination. The physical assessment consisted of a manual examination of the upper quadrant and followed accepted orthopedic PT assessment principles and techniques. 35,51-53 Specifically, the assessment included examination of the cervical spine, elbow joint complex, wrist joint, neurodynamics of the radial nerve, and the site of lesion. The cervical spine was examined by assessing all active movements through available range with the addition of over-

310 PROGNOSTIC FACTORS FOR LATERAL EPICONDYLITIS, Waugh Table 1: Descriptive and Prognostic Variables Collected From Baseline Assessment Prognostic Variables Categorization of Prognostic Variables Descriptive Variables 1. Demographic and injury characteristics 1. Demographic and injury characteristics Age Continuous Hand dominance Sex Male, female History of treatment Duration of symptoms Continuous (wk) History of sports/leisure activities Mechanism of onset Work, sport Description of symptoms History of episodes Yes, no Repetitive job ( 25h/wk performing repetitive task) vs nonrepetitive job 2. Physical findings 2. Physical findings Cervical spine articular signs Asymptomatic, symptomatic Wrist articular signs Abnormal neurodynamics No, yes Site of lesion (ECRB, ECRL, supinator, Elbow articular signs Absent, present brachioradialis) Site of lesion (ECRB) Tenoperiosteal origin, tendon body Abbreviations: ECRB, extensor carpi radialis brevis; ECRL, extensor carpi radialis longus. pressure to determine end-feel and pain provocation. Passive intervertebral movements into flexion, extension, side flexion, and rotation were then assessed at levels C4 to T1. This was followed by passive accessory intervertebral movements at these same levels, using inferior and superior glides of the facet joints. 42 Passive intervertebral and accessory movements were determined to be abnormal if the following 3 signs were present: pain was provoked, there was an abnormal quality of resistance to movement, and an abnormal end-feel was palpated. 54,55 Finally, combined movements into extension-rotation and flexion-rotation were assessed to determine end-feel and pain provocation. 52 A joint was determined to be symptomatic if at least 1 active movement was painful and exhibited an abnormal end-feel or a combined movement was painful and exhibited an abnormal end-feel, and both of the corresponding passive intervertebral movements and passive accessory intervertebral movements were abnormal as determined by the previously described decision rule. The elbow joint complex was examined by measuring active and passive flexion, extension, supination, and pronation movements with a universal goniometer. 35,56 Combined passive elbow movements, accessory movements of the radiohumeral, ulnohumeral, and superior radioulnar joints, and axial loading of the radiohumeral joint completed the elbow articular assessment. 53 Reproduction of mild pain or subtle lack of movement ( 10 ) in any or all elbow joints constituted a minor elbow joint sign and a loss of movement more than 10 or severe pain was considered a significant elbow joint sign. 31 For the purposes of analysis, mild and severe signs were combined and elbow articular signs were categorized as either absent or present. The wrist joint was similarly examined and classified for descriptive purposes. Using a goniometer, we measured active and passive wrist flexion, extension, radial deviation, and ulnar deviation and accessory glides were performed on the radiocarpal and inferior radioulnar joints. 35,53,56 The neurodynamics of the radial nerve were examined by means of the upper-limb tension test described by Butler 32 and used by Yaxley and Jull 31 in their preliminary study of lateral epicondylitis and neural tension. The test was performed with the patient supine and the neck in a neutral position. The wrist extensor complex was put on a stretch before performing the neurodynamic test to differentiate symptoms of muscle stretch from neural tension symptoms. The following components of the upper-limb tension test were added sequentially: shoulder girdle depression, internal rotation of the shoulder, elbow extension, wrist and finger flexion, thumb flexion and wrist ulnar deviation, shoulder abduction, and neck side flexion to opposite side. Subjects were classified into 3 categories according to their responses to the test. The categories were (1) normal; (2) abnormal dynamics (adverse tension) with no reproduction of lateral epicondylitis symptoms; or (3) abnormal dynamics with reproduction of lateral epicondylitis symptoms. Because the mean outcome scores for the 2 groups (normal and abnormal with no lateral epicondylitis symptom reproduction) were virtually identical, we collapsed them into 1 category for purposes of prognostic modeling and compared them with the group that had abnormal dynamics with reproduction of symptoms. Finally, the soft tissues at the elbow were examined to identify the specific site of lesion. The objectives of this part of the physical assessment were to identify the primary muscle responsible for the subject s elbow symptoms, to locate the maximum site of tenderness on the extensor carpi radialis brevis (ECRB), if that muscle was determined to be symptomatic, and to identify a secondary muscle, if present. To this end, the assessment consisted of manual muscle testing and lengthtension testing of the wrist extensors (ECRB, extensor carpi radialis longus [ECRL]), supinator, and brachioradialis, and palpation of the origin of the ECRL on the supracondylar ridge of the humerus, the tenoperiosteal junction and tendon body of the ECRB, trigger points in the wrist extensor muscle belly and supinator muscle, the annular ligament, and radiohumeral joint line. 33,35,36,56 The primary muscle was determined by maximum pain provocation with manual muscle and length-tension tests. Further differentiation between the ECRB and ECRL was dependent on pain location on palpation. This information was used primarily for descriptive purposes. However, the specific site of lesion on the ECRB (tenoperiosteal junction or tendon body) was identified by maximum point of tenderness, 26,31,36 and this dichotomous variable was used in the prognostic analysis. Eight-week follow-up. At 8 weeks, participants were mailed the DASH questionnaire and pain VAS. Phone calls were made to nonrespondents at 10 days after the mailing. Participants also completed a questionnaire on the treatment that they had received in the previous 8 weeks, including compliance with a home exercise program, activity modification, elbow brace usage, cortisone injections, nonsteroidal antiinflammatory drugs (NSAIDs) usage, and other non-pt treatments. This information was augmented by a questionnaire completed by the treating physical therapists specifying the number of treatments and types of treatment of the elbow, cervical spine, shoulder, or wrist.

PROGNOSTIC FACTORS FOR LATERAL EPICONDYLITIS, Waugh 311 Statistical Analysis Data analysis followed standard procedures for a prognostic study. 57 Summary descriptive statistics were computed on all variables. Bivariate analysis was performed to identify associations between the 9 individual predictor variables and final DASH and VAS scores and associations or colinearity between variables. Pearson correlation analysis and linear regression were used when both variables were continuous, t tests were used when 1 variable was continuous and the other dichotomous, and continuity-adjusted chi squares were calculated when both variables were categorical. The Mann Whitney U test was used for analysis of the association between dichotomous variables and final DASH and VAS scores, because these scores did not follow a Gaussian distribution. Multiple linear regression models were computed by using a forward stepwise approach and r 2 values were used as the criterion to enter variables into the model. This analysis was conducted using the PROC REG procedure in SAS, version 6.12. 58,a A variable remained in the final model if it achieved a significance level of P of.05 or less, if the r 2 statistic for the model increased by at least 10%, and if the coefficient did not change by more than 10% with the addition of other variables into the model. 57,59 Baseline scores were included in the model to adjust for the effect of baseline differences on final scores. 60 Treatment variables were analyzed to determine whether differences in treatment between participants were associated with differences in outcome. If so, we included them in the regression analysis. Specifically, cervical treatment (treated, not treated), neurodynamic mobilization (treated, not treated), modification of primary aggravating activities (stopped or modified, continued with normal activities), and number of treatments over the 8-week time period were evaluated by linear regression or by the Mann Whitney U test to determine if they were associated with final DASH or pain scores. Those treatment variables that were associated with the final outcome scores and had a P value of.50 or less were entered into the original model. Only number of treatments was associated with final pain VAS scores at this level and was included in the linear regression analysis. To limit the number of variables entered into the linear regression model, thereby preserving power and the reliability of the model, variables were screened for the strength of their association with the final DASH and pain scores. A conservative significance level of P equal to.50 or less was required for entry of a variable into the model. This was particularly relevant for the subanalysis of women (n 47). The analysis would be underpowered to include all 9 predictor variables being investigated in this study. By using this conservative screening rule, 5 variables were identified and entered into the model, thus retaining adequate power. Table 4 lists the variables included in the original models. RESULTS Participants Of the 86 participants, 3 were lost to follow-up (96.5% response rate). All analysis was therefore conducted on 83 participants. Demographic and injury characteristics. The demographic and injury characteristics of the sample are in table 2. The participants had a mean age of 44 8 years. The duration of symptoms ranged widely, from 1 week to 9 years (mean, 7mos), but the majority of participants (53%) reported a duration of between 1 and 3 months. Thirty-seven percent of participants attributed their injury to a sports activity whereas only 23% reported a work-related onset. Tennis was the sporting activity most often associated with onset of symptoms (45% of sport-related injuries) and computer usage was identified as the most common cause of work-related injuries (74% of work injuries). Of the 49% of participants who reported that they had a repetitive job, 93% reported that these repetitive job duties were computer related. Finally, although the vast majority of participants reported aching-type symptoms (81%) and/or sharp, stabbing pain (61%), a substantial group also reported experiencing pins and needles and/or numbness (27%). Physical characteristics. The physical characteristics of the sample are in table 2. A majority of participants (57%) had symptomatic cervical signs. Lateral epicondylitis symptoms were reproduced from the cervical spine in only 2 participants. Forty-one percent of participants had positive radial neural tension tests that reproduced their lateral epicondylitis symptoms. A small majority (52%) exhibited elbow articular signs, but only 7% were considered significant. The most common primary site of lesion was the tenoperiosteal junction of the ECRB. Although the supinator muscle was the primary source of symptoms in only 19% of participants, it was the most common source of secondary symptoms. In total, 55% of participants had some involvement of the supinator muscle. The ECRL signs and symptoms were present in only 6% of the sample. Conservative (PT) management. Participants received an average of 10 treatment sessions over the 8 weeks. All participants were treated with at least 1 type of physical modality. Ultrasound was the most frequently used modality and was provided to 81% of patients. Deep transverse friction massage was the most common type of manual treatment, with 71% of participants being treated with this technique. All but 2 participants were instructed in an exercise program that consisted most frequently of a combined stretching and strengthening program for the wrist extensor muscles. Thirty-seven percent of the participants received treatment for the cervical spine or shoulder in addition to local elbow treatment. Eighteen percent of participants used NSAIDs and 4 subjects received cortisone injections. Based on this information, it is apparent that the typical PT management strategy provided to the participants in this study was a combination of ultrasound, deep transverse friction massage, and a stretching and strengthening program for the wrist extensor muscles. Prediction Model: DASH Scores Bivariate results. A summary of the DASH scores for the sample is outlined in table 3. The sample as a whole improved over the 8-week treatment period: follow-up DASH scores were significantly lower than baseline scores as determined by the Wilcoxon signed-rank test (P.0001). Sex, mechanism of onset (work vs sport), repetitive jobs, cervical joint signs, abnormal neurodynamics, and self-reported nerve symptoms (pins and needles and/or numbness) were individually associated with final DASH scores at a P level of.05 or less. Each of these variables, with the exception of self-reported nerve symptoms, was also associated with high baseline scores. Consequently, if the factor was present at baseline, it was not only associated with higher baseline scores but also predictive of a higher follow-up score (fig 1). This occurred because participants who presented with these prognostic factors experienced the same amount of change in DASH scores over the 8-week time period as those who did not have these factors, despite the fact that high baseline scores would allow for more change than low scores. Baseline scores

312 PROGNOSTIC FACTORS FOR LATERAL EPICONDYLITIS, Waugh Table 2: Demographic, Injury, and Physical Characteristics of Entire Lateral Epicondylitis Sample and Subdivided by Sex Characteristics n (N 83) % of Total Sample Women (n 47) Men (n 36) Total (N 83) Mean age SD (y) 43.2 7.7 45.5 8.7 44.2 8.2 Mean duration SD (wk) 39.5 81.5 19.9 25.9 31.0 64.1 Hand dominance (side of symptoms) Dominant 68 81.9 41 27 Mechanism of onset Insidious 14 16.9 8 6 Work 19 22.9 18 1 Sports 31 37.3 14 17 Other* 19 22.9 8 11 Repetitive job duties 25h/wk 42 51.0 28 13 Elbow symptoms Aching 67 80.7 43 24 Sharp, stabbing 51 61.4 25 26 Numbness, pins and needles 22 26.5 15 7 Burning 18 21.7 9 9 History of previous episode(s) 24 30.0 15 9 Cervical spine articular signs 47 56.7 32 15 Abnormal neurodynamics: radial nerve bias (with reproduction of epicondylitis symptoms) 34 59.0 23 11 Elbow articular signs 43 51.8 28 15 Wrist articular signs 13 15.7 7 6 Site of lesion: primary muscle ECRB 67 80.7 40 27 Supinator 16 19.3 7 9 Site of lesion: secondary muscle Supinator 30 36.0 22 8 ECRB 26 31.3 12 14 ECRL, brachioradialis 6 7.2 3 3 Abbreviation: SD, standard deviation. *Activities such as moving furniture, carrying luggage, and painting a room were performed only once but precipitated onset of symptoms. Percentages exceed 100% because subjects experienced more than 1 symptom. did, therefore, correlate with final DASH scores (r.68, P.0001). Multiple regression results. Because only 1 man reported a work-related injury, the mechanism of onset variable could not be used in regression analysis. It was replaced with the repetitive jobs variable, which was associated with work injuries but was also present among men, and therefore applicable to the entire sample. Although the variable of self-reported nerve symptoms was not one of the our hypothesized prognostic factors, we collected it for descriptive purposes. However, because results of the bivariate analysis indicated that it was not only associated with final DASH scores but was also unrelated to the other variables being investigated, we included it in the multiple regression analysis. Table 4 shows the forward stepwise regression results for the full model. The final model includes the baseline score, sex, and self-reported nerve symptoms (table 5), and explains 61% Table 3: Performance of the Outcome Measures in Lateral Epicondylitis Sample Baseline 8 Weeks Change Score 95% CI for Change Score DASH 33.0 15.9 19.4 16.2 13.6 13.5 10.6 16.3 Pain VAS 45.0 22.8 20.1 22.2 24.9 27.4 18.8 31.3 NOTE. Values are mean SD. Abbreviation: CI, confidence interval. Fig 1. Sex as an example of the effect of a significant factor on DASH scores over the 8-week study period: women are associated with both higher baseline scores and final scores than men, with a similar amount of change occurring between baseline and followup.

PROGNOSTIC FACTORS FOR LATERAL EPICONDYLITIS, Waugh 313 Table 4: Full Model From Forward Stepwise Multiple Linear Regression Analysis: Entire Sample Variable Coefficient 95% CI P Value R 2 Final DASH score Baseline score (cont) 0.62 0.44 0.80.01.46 Sex (male/female) 10.62 5.02 16.22.01.57 Nerve symptoms (yes/no) 7.32 0.76 13.88.03.62 Repetitive job (yes/no) 4.22 1.20 to 9.64.13.63 Duration of symptoms (cont) 0.12 0.28 to 0.04.14.65 Age (cont) 0.21 0.13 to 0.55.07.66 Elbow joint signs (yes/no) 2.17 3.09 to 7.43.41.66 Neurodynamics (yes/no) 0.73 6.61 to 5.15.81.66 Cervical joint signs (yes/no) 0.61 5.57 to 6.79.86.66 Final pain VAS score Nerve symptoms (yes/no) 18.81 8.41 29.21.01.19 Sex (male/female) 9.28 0.13 18.43.04.25 Baseline score (cont) 0.19 0.01 0.27.05.31 Number of treatments (cont) 0.71 1.57 to 0.15.10.34 Repetitive job (yes/no) 5.30 3.51 to 14.12.23.36 Age (cont) 0.29 0.21 to 0.79.26.36 Elbow joint signs (yes/no) 3.47 5.31 to 12.25.43.37 Neurodynamics (yes/no) 2.65 12.73 to 7.40.60.37 Site of lesion (origin/tendon body) 1.29 6.75 to 4.17.63.38 Cervical joint signs (yes/no) 3.75 5.55 to 13.07.42.38 Abbreviation: cont, continuous. *Results are based on a forward stepwise approach; the values in the table were adjusted for each variable already entered into the model. of the total variability of the final DASH scores. This model indicates that, for example, if baseline scores were identical, a woman with nerve symptoms would have a final DASH score that would be 16 points higher than a man with no report of nerve symptoms. Prediction Model: Pain VAS Scores Bivariate results. A summary of pain VAS scores for the sample is provided in table 3. The sample as a whole improved over the 8-week treatment period: follow-up pain scores were significantly lower than baseline scores as determined by the Wilcoxon signed-rank test (P.001). Sex, mechanism of onset (work vs sport), repetitive jobs, and self-reported nerve symptoms were individually associated with final pain scores at a P level of.05 or less. Unlike DASH scores, baseline pain scores did not correlate with final scores (r.18, P.12). Participants with these prognostic factors presented with similar baseline scores but experienced less change over the 8-week treatment period, so that at, follow-up, these participants had significantly higher pain scores than participants without these factors (see fig 2 for an example). Multiple regression results. For the reasons noted previously, variables for repetitive jobs and self-reported nerve symptoms were included in the regression analysis. Table 4 shows the forward stepwise regression results for the full model. The final model includes the baseline score, sex, and self-reported nerve symptoms (see table 5) and explains 31% of the total variability of the final VAS scores. This model indicates that, for example, if baseline scores were identical, a woman with nerve symptoms would have a final VAS score that would be 24.0mm higher than a man without nerve symptoms. Subanalysis: Women To determine reasons for the disparity in prognosis between men and women, additional analysis was performed on the subset of women (n 47). Results showed that being female was associated with work-related onsets (only 1 man presented Table 5: Final Prediction Models for 8-Week DASH and Pain VAS Scores: Entire Sample Dependent Variable Independent Variables Coefficient 95% CI Final DASH Final VAS Adjusted R 2 for the Model MSE for the Model P Value for the Model Intercept 5.77 0.07 to 11.61.61 9.55.0001 Baseline DASH score (per unit) 0.50 0.34 0.66 Sex (female) 8.92 3.32 14.52 Nerve symptoms 7.32 0.80 13.84 Intercept 0.57 11.03 to 9.89.31 15.58.0003 Baseline VAS score (per unit) 0.19 0.01 0.37 Sex (female) 9.29 0.35 18.23 Nerve symptoms 15.08 4.67 25.50 Abbreviation: MSE, mean standard error.

314 PROGNOSTIC FACTORS FOR LATERAL EPICONDYLITIS, Waugh Fig 2. Sex as an example of the effect of a significant factor on pain VAS scores over the 8-week study period: women and men have similar baseline scores but the amount of change over the 8 weeks is lower for women, resulting in significantly higher follow-up scores for women. with a work onset), repetitive jobs (relative risk [RR] 1.7; 95% confidence interval [CI], 1.2 2.4) and cervical joint signs (RR 1.6; 95% CI, 1.2 2.7) and that these factors had significant prognostic influence on their recovery. Multiple linear regression analysis conducted on the final DASH scores of women generated a prediction model with the factors cervical joint signs and repetitive jobs (table 6). This model predicts that a woman with a repetitive job and cervical joint signs will have a final DASH score 21.0 points higher than a woman with a nonrepetitive job and normal cervical signs. Multiple regression analysis performed on the final pain VAS scores of women produced a model with the factor mechanism of onset (see table 4). This model indicates that a woman who attributes her injury to work will have a final pain score that is 21.4mm higher than a woman who attributes her injury to a sporting activity. DISCUSSION This study showed that women and self-reported nerve symptoms were associated with higher DASH and VAS scores after 8 weeks of PT management for lateral epicondylitis. Of the 2 previous studies that specifically investigated sex as a prognostic factor, Gerberich and Priest 23 also found that significantly more men reported complete recovery than women. This is in contrast to Stratford et al 26 who found no difference in recovery between men and women. Stratford measured outcome as a dichotomy of success or failure based on the ability of the subject to progress to an exercise program as determined by a blinded assessor. These conflicting results, then, may be because of interstudy variability with respect to the outcome measures used: self-report versus performance measures. Although Gerberich and Priest 23 reported a prognostic difference between men and women, they did not provide any explanation for this disparity. The results of the subanalysis conducted in the present study may assist in offering such an explanation. Much of the difference in prognosis between men and women in the present study can be explained by the fact that women presented with different characteristics than men and that, for women, these characteristics were associated with a poorer prognosis. Specifically, as noted, women were more likely than men to have work-related onsets, repetitive jobs, and/or positive cervical signs. Moreover, work-related onsets were associated with both repetitive jobs (RR 2.5; 95% CI, 1.5 4.2) and cervical signs (RR 2.1; 95% CI, 1.2 2.9) and, therefore, these factors were likely to coexist. Indeed, the largest subgroup of women (36%) presented with all 3 of these prognostic factors and had a mean final DASH score of 31.3 17.9 and VAS score of 38.1 29.7. Conversely, the largest subgroup of men (35%) presented with sport-related onsets, nonrepetitive jobs, and negative cervical signs, all factors associated with a better prognosis, and had a mean final DASH score of 9.7 6.6 and VAS score of 14.0 14.4. A further investigation of the effect of cervical signs on prognosis provided additional insight into the differences between men and women. Seventy-nine percent of women in repetitive jobs also had concurrent cervical signs whereas only 46% of men in repetitive jobs presented with cervical signs. Among women, the influence of cervical signs on prognosis was magnified when it occurred in combination with repetitive jobs. This interaction was not statistically significant, but the subgroup of women in repetitive jobs without cervical signs was very small (n 6), which limited power to detect an interaction. The effect of cervical dysfunction on prognosis on its own and in conjunction with repetitive jobs was not observed in men. Men, therefore, were less likely than women to present with cervical signs and, when present, this factor did not exhibit the same influence on prognosis as that observed in women. Analysis of the data did not clarify this incongruity. It may be that work-related onsets, repetitive jobs, and cervical dysfunction also have a prognostic influence on men. However, inadequate statistical power in the present sample might have precluded detection of an association between these factors and prognosis for men: the sample included only 1 man with a work-related onset and 6 men with both repetitive jobs and positive cervical signs. Further, other considerations that could help explain the lack of association between the cervical spine Table 6: Final Prediction Model for 8-Week DASH and Pain VAS Scores: Female Subgroup (n 47) Dependent Variable Independent Variables Coefficient 95% CI Final DASH Final VAS Adjusted R 2 for the Model MSE for the Model P Value for the Model Intercept 9.02 0.34 to 18.38.23 13.70.008 Cervical signs 11.70 1.70 21.7 Repetitive job 9.15 0.05 18.3 Intercept 13.57 1.49 25.65.20 22.60.01 Mechanism of onset (work) 21.43 4.98 37.88

PROGNOSTIC FACTORS FOR LATERAL EPICONDYLITIS, Waugh 315 and repetitive jobs among men, such as workstation ergonomics and anthropometric data, were not collected as part of this prognostic study. Mechanism of onset has been reported as related to prognosis in lateral epicondylitis subjects. Akermark et al 25 noted a pronounced difference in outcome between work- and sportsrelated cases in their RCT. They found that, even in their placebo group, spontaneous recovery was unexpectedly high in the sport cases. They suggest that this recovery could be due to the ease of avoiding aggravating sporting activities, compared with the difficulty of avoiding aggravating work activities. Results from the present study showed that subjects who had a sport onset were no more likely to modify their aggravating activities than those with a work injury. However, the work group reported that their aggravating activity was actually a leisure or sport activity, not a work duty. Only 2 subjects were on modified work duties and, therefore, the continued participation in normal job duties may, indeed, be partly responsible for the prognostic differences observed between the work and sport cases. The repetitive type job engaged in by the majority of subjects in the work group, the high incidence of cervical signs in subjects with work-related cases, and the amplified effect of the combination of repetitive job duties and cervical signs on outcome may each provide additional reasons for the prognostic influence of work-related onsets. Ninety-three percent of the repetitive job duties described by the participants in this sample involved the use of computer monitor, keyboard, and mouse. The relation between repetitive keyboarding jobs and cervical signs and their combined negative impact on the prognosis of lateral epicondylitis have not been previously identified. Other authors 17,19,27-30,61,62 have suggested that cervical spine pathology may contribute to a poor prognosis of lateral epicondylitis. Gunn and Milbrandt 27 postulated that lateral epicondylitis symptoms could be the direct result of a reflex localization of pain from radiculopathy of the cervical spine. Yet, this does not fully explain why cervical signs have a greater impact on the prognosis of participants with repetitive keyboarding duties than those engaged in nonrepetitive jobs. One could postulate that the underlying cause might be related to prolonged sitting postures. It may be that few people have adequate endurance of the cervical stabilization muscles to sustain a balanced neck position for the 5 hours or more per day required for these repetitive jobs. As the muscles fatigue, segments of the lower cervical spine translate anteriorly, leading to a forward head posture. According to Lee, 30 this type of posture typically causes the cervicothoracic junction to become hypomobile and the C6 segment to become hypermobile. Hypermobility of this segment could then compromise the C6 nerve root, because of friction or traction forces. 30,63 The resultant ischemia and denervation could subsequently cause painful trigger points in the forearm musculature, weakness of the wrist extensors, and formation of inadequate scar tissue, leading to protracted lateral epicondylitis symptoms. It might be expected that abnormal neurodynamics of the radial nerve would accompany this type of scenario. 32 Indeed, in the present sample, cervical signs were significantly associated with a positive upper-limb tension test that reproduced the participants elbow symptoms (RR 2.5; 95% CI, 1.3 4.8). Even taking into account these other factors, women still had higher final DASH and pain scores than did men. This sex difference in the self-report of disability and pain has been found in other populations. For example, women consistently report higher levels of disability and pain associated with osteoarthritis and rheumatoid arthritis. 64-67 Various explanations have been provided for these selfreport differences: women are socialized to acknowledge pain and discomfort whereas men are socialized to be stoic and self-reliant 68,69 ; men and women use different coping strategies 64,65 ; and women may recognize and interpret symptoms more readily than men. 68 One could speculate that other reasons specific to the lateral epicondylitis population may include activities such as household chores and child-rearing duties that may be predominately carried out by women in addition to their work activities, thereby putting additional demands on their forearm muscles. However, this information was not collected as part of the present study. The second factor related to a poor prognosis for the entire sample was the self-report of symptoms of pins and needles and/or numbness. These symptoms, which may indicate nerve involvement, were not statistically associated with a positive upper-limb tension test or cervical signs, although 78% of participants with these symptoms did present with the latter. The participants also did not present with the characteristic signs of radial tunnel syndrome (RTS) deep aching throughout the forearm musculature and pronounced pain with resisted third finger extension and forearm supination. 70-74 Moreover, it is rare to have paresthesia with RTS, because the posterior interosseous nerve does not contain cutaneous sensory fibers. 75 Further investigation that includes a complete neurologic examination and nerve conduction tests is needed to determine the pathology related to these symptoms and to help clarify the reason for their influence on prognosis. Five of the potential prognostic factors we investigated did not significantly relate to the 8-week outcome: site of lesion, history of prior episodes, duration of symptoms, age, and elbow articular symptoms. Stratford et al 76 reported that a lesion at the origin of the ECRB was associated with a poorer outcome than a lesion in the tendon body itself. The present study did not reproduce these findings. The present study also found that the origin of the ECRB was the primary site in 60% of participants, which is considered typical for this population. 4,12,36,77,78 Stratford 76 reported that the origin was the primary site in only 20% of participants. Poor reliability of assessment techniques may have led to these conflicting results. Stratford et al 26 also reported that a history of prior episodes was associated with poor outcome. The present study found that a prior history of symptoms was associated with higher initial DASH scores, but at 8 weeks this factor was no longer statistically significant. The Stratford study, which was conducted over a shorter period ( 5 wk), might well have come to a conclusion consistent with the present study if the patients had been followed longer. A history of previous episodes may in fact require a slightly longer period of rehabilitation, but not longer than 8 weeks. The prognostic study by Gerberich and Priest 23 reported that a mean duration of symptoms greater than 17.3 months correlated with a self-report of no recovery and a mean of 6.3 months correlated with complete recovery after conservative treatment. These results, however, were based on their bivariate analysis. Symptom duration did not remain in their final model after multiple regression analysis. Solveborn et al 18 also noted in their study investigating the effectiveness of cortisone injection with anesthetic additives that a history of epicondylalgia longer than 3 months in duration was associated with a poorer prognosis as measured by a pain VAS. No such result was observed in the present prognostic study. This difference may be because of the different types of treatment provided in the 2 studies PT versus steroid injection. Nevertheless, fur-

316 PROGNOSTIC FACTORS FOR LATERAL EPICONDYLITIS, Waugh ther research is required to elucidate the effect of duration of symptoms on prognosis. Age has not been previously found to be of prognostic significance, 23,26 and the present study further confirms this conclusion. The impact of elbow articular signs on prognosis of lateral epicondylitis has not been previously investigated, and we did not find that this factor was related to outcome after 8 weeks of conservative treatment. This could imply that elbow articular dysfunction responds well to standard PT management of lateral epicondylitis. Limitations. The present study has several limitations. The first is the lack of generalizability of the results, because most of the sample was recruited from private sports medicine clinics and 96% of the participants had white-collar jobs. Also, one cannot determine how representative the sample is of the population from which it was recruited because information about the number and characteristics of patients who were excluded from the study or who refused to participate was inadequate. A second limitation is the absence of confirmed intra- and interrater reliability of the assessment techniques used in the baseline clinical examination, although moderate evidence exists of the reliability of a manual examination of the cervical spine 54,55,79 and shoulder. 80,81 Because the subanalysis of women was exploratory and had a small sample size, only a limited number of variables could be considered. These factors could affect the validity and generalizability of the results for this subgroup. Significance of results. The results of the present study are relevant to both clinicians and researchers. First, they will help physical therapists identify patients who may experience high disability and pain levels after 8 weeks of treatment, based on a standard PT clinical examination performed in a clinical setting. For example, a woman presenting with positive cervical signs and engaged in a repetitive keyboarding job would be predicted to have a poorer outcome than a man with a sportsrelated onset and a nonrepetitive job. Second, the results of the present study may help direct both the assessment and management of patients with this syndrome. The knowledge that cervical signs are frequently present in a patient with a repetitive keyboarding job should direct the treating practitioner to include the cervical spine in the assessment and ask appropriate questions regarding work duties and mechanism of onset. A management strategy that includes cervical spine treatment and modification of the patient s work-site and work duties might help lessen the effect that work-related onsets, repetitive jobs, and cervical signs have on outcomes for women with lateral epicondylitis. Third, the results of this study may help improve the specificity of inclusion criteria for future clinical trials and promote the evaluation of therapeutic interventions on selected homogeneous groups of lateral epicondylitis patients, such as on women with repetitive keyboarding jobs. Future directions. Because the manual examination techniques have reliability issues and because many factors in the present study were investigated for the first time, a similar study must be conducted to determine if our results can be reproduced in this and other populations. Other populations should include workers compensation cases and blue-collar workers receiving PT through the publicly funded health care system. In particular, self-reported nerve symptoms and repetitive keyboarding jobs, which were not included in the a priori hypothesis of this study, must be reevaluated. Further investigation is needed into reasons for the prognostic disparity between women and men and into the relationship between cervical articular signs, repetitive keyboarding jobs, and lateral epicondylitis. Additional research is required to determine whether work-related onsets, repetitive jobs, and cervical signs also have a prognostic influence on men, and, if not, to identify why. Finally, because symptoms often persist for longer than 8 weeks, 6,82,83 a prognostic study with a longer follow-up must be conducted to identify factors associated with chronicity. The present study is currently being extended to achieve this objective. CONCLUSIONS Certain subgroups of participants who presented with specific demographic, injury, and physical characteristics experienced a poorer outcome after 8 weeks of PT management, as measured by disability and pain levels. In particular, participants who reported symptoms of pins and needles and/or numbness had higher DASH and pain VAS scores than those who reported the more common symptoms of aching or sharp pain. Women had a poorer outcome than did men. This was partly because women were more likely than men to have work-related onsets, repetitive keyboarding jobs, and positive cervical signs, all factors associated with higher DASH and VAS scores. Conversely, men were more likely to have a sports-related injury, to be engaged in nonrepetitive jobs, and to have normal cervical findings, all factors associated with lower DASH and VAS scores. We anticipate that the results of this study will advance the understanding of the differences that exist among people suffering from lateral epicondylitis and the influence that these differences can have on prognosis. References 1. Allander E. Prevalence, incidence and remission rates of some common rheumatic diseases and syndromes. Scand J Rheumatol 1974;3:145-53. 2. Gruchow HW, Pelletier D. An epidemiologic study of tennis elbow. Incidence, recurrence, and effectiveness of prevention strategies. Am J Sports Med 1979;7:234-8. 3. Maylack FH. Epidemiology of tennis, squash and racquetball injuries. Clin Sports Med 1988;7:233-43. 4. Nirschl RP. Elbow tendinosis/tennis elbow. Clin Sports Med 1992;11:851-70. 5. Wadsworth TG. Tennis elbow: conservative, surgical, and manipulative treatment. BMJ (Clin Res Ed) 1987;294:621-4. 6. Coonrad RW, Hooper WR. Tennis elbow: its course, natural history, conservative and surgical management. J Bone Joint Surg Am 1973;55:1177-82. 7. Dimberg L. The prevalence and causation of tennis elbow (lateral humeral epicondylitis) in a population of workers in an engineering industry. Ergonomics 1987;30:573-80. 8. Kivi P. The etiology and conservative treatment of humeral epicondylitis. Scand J Rehabil Med 1983;15:37-41. 9. Binder AI, Hazleman BL. Lateral humeral epicondylitis a study of natural history and the effect of conservative therapy. Br J Rheumatol 1983;22:73-6. 10. Chard MD, Hazleman B. Tennis elbow a reappraisal. Br J Rheumatol 1987;28:186-8. 11. Verhaar JA. Tennis elbow. Anatomical, epidemiological and therapeutic aspects. Int Orthop 1994;18:263-7. 12. Nirschl RP. The etiology and treatment of tennis elbow. J Sports Med 1974;2:308-23. 13. Rettig AC. Elbow, forearm and wrist injuries in the athlete. Sports Med 1998;25:115-30. 14. Ollivierre CO, Nirschl RP. Tennis elbow. Current concepts of treatment and rehabilitation. Sports Med 1996;22:133-9. Comment in: Sports Med 1998;25:137. 15. Behr CT, Altchek DW. The elbow. Clin Sports Med 1997;16:681-704. 16. Smidt N, van der Windt DA, Assendelft WJ, Deville WL, Korthals-de Bos IB. Corticosteroid injections, physiotherapy, or a wait-and-see policy for lateral epicondylitis: a randomised controlled trial. Lancet 2002;359:657-62.