Damage to the anterior cruciate ligament (ACL) is a major. Clinical Diagnosis of an Anterior Cruciate Ligament Rupture: A Meta-analysis

Transcription

1 Clinical Diagnosis of an Anterior Cruciate Ligament Rupture: A Meta-analysis Anne Benjaminse, PT 1 Alli Gokeler, PT 2 Cees P. van der Schans, PT, PhD 3 Journal of Orthopaedic & Sports Physical Therapy Study Design: Meta-analysis. Objectives: To define the accuracy of clinical tests for assessing anterior cruciate ligament (ACL) ruptures. Background: The cruciate ligaments, and especially the ACL, are among the most commonly injured structures of the knee. Given the increasing injury prevalence, there is undoubtedly a growing need for clinical decision making of health care providers. We reviewed the literature to analyze the diagnostic accuracy of the clinical examination for assessing ACL ruptures. Methods and Measures: MEDLINE (1966 to April 2005), EMBASE (1989 to April 2005), and CINAHL (1982 to April 2005) searches were performed. Also reference lists of the included studies were reviewed. Studies selected for data extraction were those that addressed the accuracy of at least 1 physical diagnostic test for ACL rupture and compared the performance of the clinical examination of the knee with a reference standard, such as arthroscopy, arthrotomy, or MRI. Searching was limited to English, German, and Dutch languages. Results: Twenty-eight studies that assessed the accuracy of clinical tests for diagnosing ACL ruptures met the inclusion criteria. Study results were, however, heterogeneous. The Lachman test is the most valid test to determine ACL tears, showing a pooled sensitivity of 85% (95% confidence interval [CI], 83-87) and a pooled specificity of 94% (95% CI, 92-95). The pivot shift test is very specific, namely 98% (95% CI, 96-99), but has a poor sensitivity of 24% (95% CI, 21-27). The anterior drawer test shows good sensitivity and specificity in chronic conditions, respectively 92% (95% CI, 88-95) and 91% (95% CI, 87-94), but not in acute conditions. Conclusion: In case of suspected ACL injury it is recommended to perform the Lachman test. Because the pivot shift test is very specific both in acute as well as in chronic conditions, it is recommended to perform the pivot shift test as well. J Orthop Sports Phys Ther 2006;36(5): doi: /jospt Key Words: accuracy, anterior cruciate ligament, examination, knee Damage to the anterior cruciate ligament (ACL) is a major injury to the knee joint, 18,50,68,72 with increasing prevalence during sporting activities. 47 For example, according to Griffin et al, 24 an estimated ACL tears occur annually in the United States, with requiring surgical reconstruction. With an estimated cost for these injuries of almost a billion dollars per year, identifying risk factors and developing 1 Physical Therapist (at time of study), Department of Physical Therapy, Medisch Centrum Zuid, Groningen, The Netherlands; Junior Researcher (at time of study), Hanze University Groningen, Center for Research and Development in Health Care and Nursing, Groningen, The Netherlands. 2 Sports Physical Therapist, Researcher, University Medical Center Groningen, Center for Rehabilitation Groningen, The Netherlands. 3 Professor in Health Care and in Nursing, Hanze University Groningen, Center for Research and Development in Health Care and Nursing, Groningen, The Netherlands. Address correspondence to Cees P. van der Schans, Hanze University Groningen, Center for Research and Development in Health Care and Nursing, Eyssoniusplein 18, 9714 CE Groningen, The Netherlands. c.p.van.der.schans@pl.hanze.nl prevention strategies and optimal treatment methods have widespread health and economical implications. 24 ACL rupture may leave the patient with knee joint instability during functional activities. In the absence of the ACL the task of restraint falls upon secondary ligamentous structures. So, with time, secondary restraints of the ACLdeficient knee may also become lax or injured. Surgical reconstruction of the ACL is the preferred treatment in active patients presenting with instability. 11 The longterm effect of instability may increase the probability of early onset of osteoarthritis. 51 Based on a recent literature meta-analysis, ACL reconstruction seems to be superior to conservative treatment to resolve objective and subjective joint instability and help the return to preinjury level of sports. 29 From this perspective, it is of fundamental importance to perform the most accurate tests available to diagnose an ACL rupture, particularly for physical therapists that practice in a sports and orthopaedic setting without a referral by a physician. They rely heavily on clinical tests and therefore need to be able to screen for ACL injury to make appropriate treatment decisions and to make referral for further imaging and diagnosis and possible surgery. 22 Three commonly applied tests in clinical practice to determine LITERATURE REVIEW Journal of Orthopaedic & Sports Physical Therapy 267

2 Potential studies identified and screened for retrieval (n = 7143) Full text of the studies retrieved for more detailed evaluation (n = 48) Studies included in metaanalysis, including ref 55 (n = 28) FIGURE 1. Flow diagram of search strategy. Studies excluded because considered not potentially relevant (n = 7095) Studies not included with reasons (n = 21) ACL injuries are the anterior drawer, the Lachman, and the pivot shift tests. For the anterior drawer test, the patient is supine with the hip flexed to 45, the knee flexed to 90, and the lower leg in neutral rotation. The examiner stabilizes the patient s foot with his thigh and places both hands behind the proximal tibia with the thumbs on the tibial plateau. The examiner applies an anteriorly directed force to the proximal tibia and judges the amount of tibial translation. According to the International Knee Documentation Committee (IKDC 2000), 2 the anterior drawer test is rated as normal (0 to 2 mm), nearly normal (3 to 5 mm), abnormal (6 to 10 mm), or severely abnormal ( 10 mm), based on the amount of greater tibial translation on the injured side compared to the uninjured contralateral knee. Increased anterior tibial displacement in side-to-side comparison is indicative of an ACL tear. The Lachman test is performed with the patient lying supine and with the involved extremity on the side of the examiner. The femur is stabilized with 1 hand, with the patient s knee joint in 20 to 30 of flexion. The examiner s other hand is applied to the posterior aspect of the proximal tibia. An anteriorly directed force is applied to displace the tibia. Increased anterior tibial translation with a soft end point compared to the other side constitutes a positive test, indicating disruption of the ACL. As described by the IKDC 2000, 2 severity is graded as normal ( 1 to 2 mm), nearly normal (3 to 5 mm), abnormal (6 to 10 mm), or severely abnormal ( 10 mm), based on the amount of greater tibial translation on the injured side compared to the uninjured contralateral knee. The pivot shift test is performed with the patient supine. The extended leg is picked up at the ankle with the examiner s ipsilateral hand. This hand internally rotates the knee and flexes the knee from full extension, while applying valgus stress with the contralateral hand on the lateral side of the proximal tibia. A sudden reduction of the anteriorly subluxed lateral tibial plateau caused by the iliotibial tract indicates a positive pivot shift test. 23 The IKDC 2000 knee examination form grades the pivot shift test as equal, glide (+), clunk (++), or gross (+++). 2 Because data concerning the validity of clinical diagnostic tests for ACL ruptures are heterogeneous, it is difficult to conclude which test, or which combination of tests, is most appropriate for the diagnosis of suspected ACL rupture. The purpose of this meta-analysis was to evaluate the diagnostic accuracy of the anterior drawer, Lachman, and the pivot shift tests. METHODS Literature Search We performed a computerized literature search to retrieve articles pertaining to the clinical examination of patients with suspected ligamentous injury of the knee. We conducted 3 independent searches in MEDLINE (1966 to April 2005), EMBASE (1989 to April 2005), and CINAHL (1982 to April 2005). The results of these 3 searches were combined and duplicates were removed. Furthermore, studies obtained via other methods were also considered (eg, through personal contacts and screening the references cited in appropriate retrieved articles). The search strategy is shown in Figure 1 and operationally defined in Table 1. Literature Selection Each citation (title and abstract) initially identified through the search strategies was reviewed by the first author (A.B.) for potential relevance, as shown in Figure 1. Full papers of the citations identified as being possibly eligible for inclusion were obtained and independently reviewed for final inclusion (A.B., A.G., and C.S.). Minimal requirements for inclusion were as follows: (1) in vivo, human studies; (2) data reported on the accuracy of at least 1 physical diagnostic test for ACL rupture; (3) arthroscopy, 268 J Orthop Sports Phys Ther Volume 36 Number 5 May 2006

3 TABLE 1. Operationalization of the search strategy. Number MEDLINE (1966 to April 2005) EMBASE (1989 to April 2005) 1 knee injuries [MeSH] explode knee-cruciate-ligament /all subheadings 2 ligaments, articular [MeSH] explode knee-ligament-injury /all subheadings CINAHL (1982 to April 2005) explode knee injuries /all subheadings + all age subheadings explode ligaments /all subheadings + all age subheadings 3 diagnos* [TI, AB] explode knee-injury /all subheadings explode ligament-injuries /all subheadings + all age subheadings 4 diagnosis [MeSH] explode anterior-cruciate-ligamentrupture /all subheadings explode ligaments-articular /all subheadings + all age subheadings 5 diagnosis [subheading] diagnos* in ti, ab diagnos* in ti, ab 6 sensitiv* [TI, AB] explode diagnosis /all subheadings explode diagnosis /all subheadings + all age subheadings 7 specific* [TI, AB] sensitiv* in ti, ab sensitiv* in ti, ab 8 sensitivity and specificity [MeSH] specific* in ti, ab specific* in ti, ab 9 #1 AND #2 AND (#3 OR #4 OR #5 OR #6 OR #7 OR #8) explode sensitivity-and-specificity /all subheadings 10 (#1OR#2OR#3OR#4)AND(#5 OR #6 OR #7 OR #8 OR #9) arthrotomy, or magnetic resonance imaging (MRI) used as the gold standard for measuring the accuracy of the physical examination; (4) the possibility of constructing a 2 2 table specifying the true and false-positive and the true and false-negative values; and (5) written in English, German, or Dutch. Validity Assessment and Study Characteristics A standardized assessment of the methodological quality regarding the validity of the studies and the applicability of the results was adapted from the Cochrane Methods Group 1 and applied by 2 of the authors (A.B. and A.G.) to each included study. They independently assessed the methods of data collection, patient selection, blinding, and prevention of verification bias (ie, when the results of a diagnostic test affect whether the gold-standard procedure is used to verify the test result), study design, and description of the index test and reference standard. Specific questions used to evaluate these criteria are presented in the Appendix. Answers to these questions are presented in Table 2. The reviewers agreed on the answers to most of these questions. Disagreements were resolved by consensus with the assistance of a third reviewer (C.S.). Data Abstraction and Quantitative Data Synthesis Whenever possible, data were abstracted from each article by the first author (A.B.) to allow for calculation of the sensitivity and specificity of each physical explode sensitivity-and-specificity /all subheadings + all age subheadings (#1) AND (#2 OR #3 OR #4) AND (#5 OR #6 OR #7 OR #8 OR #9) examination finding. Next, the data were independently reviewed by a second reviewer (C.S.). Statistical analysis was performed with Meta-DiSc, Version 1.0.9, from Zamora et al, 75 afterconstructing2 2tables. We did not count the number of included patients; instead, we made the methodological choice to consider the number of knees with a ruptured ACL. In the evaluation of diagnostic tests, each individual study was summarized by a pair of statistics, sensitivity and specificity, that measured the test s accuracy. Then, the overall test accuracy indices were calculated as the weighted average of the summary statistics, in which the weight of each study is its sample size. MetaDiSc uses the most straightforward method to pool sensitivities and specificities. Pooled sensitivity was calculated as the total number of true positives in all studies, divided by the total number of subjects in all studies; pooled specificity was calculated as the total number of true negatives divided by the total number of subjects in all studies. This is equivalent to saying that the method approximates the inverse-variance pooling method (ie, weighting each study according to its sample size). 75 For sensitivity and specificity, the pooling method is a fixed-effect model. For the positive (LR+), the negative likelihood ratio (LR ), and the diagnostic odds ratio (DOR), the default method used is a randomeffect model. The chi-square test has been provided to assess the homogeneity of the sensitivities and specificities from studies as both measures are simple proportions. 16 We calculated the LR+ and the LR, which describe the discriminatory properties of posi- LITERATURE REVIEW J Orthop Sports Phys Ther Volume 36 Number 5 May

4 TABLE 2. Methodological quality of 28 studies. Journal of Orthopaedic & Sports Physical Therapy Criteria for Study Validity* Reference A1 A2 A3 A4 A5 A6 B1 B2 B3 B4 B5 C1 C2 C3 C4 C5 Anderson et al, 1989 AS/AT b b d a c b a a b a a a b b b Boeree et al, 1991 MRI c e c a b b a a b b b b c c a Bomberg et al, 1990 AS c b a a a c a d 1 a b b c c b Braunstein, 1982 AT b e e a c c c d b b a b b c b Cooperman et al, 1990 AS/AT b e f a b d b a 1 a a a a c a Dahlstedt et al, 1989 AS c c c a c b b d b a a b b b b DeHaven, 1980 AS c b a a a c a b 1 b b b b c b Donaldson et al, 1985 AT c b d b c b a d 2 a b a c b b Hardaker et al, 1990 AS c b c c b c a d 2 b a b b c b Harilainen et al, 1987 AS/AT b b e a b b c d 2 a a b b a b Hughston et al, 1976 AT c b c a c c a a b a a a c b b Jonnson et al, 1982 AS/AT b b c c c c a d b b b a c b b Katz et al, 1986 AS c b c b b b a d b a a a c c b Kim et al, 1995 AS c b d b c c a d 2 a b a b b b Learmonth et al, 1991 AS b b d a a b a c a b b b c c b Lee et al, 1988 MRI c a e a c b c d b b b a b b b Lucie et al, 1984 AS/AT a b d a a d a a a b b b b b b Liu et al, 1995 AS c b d b c c c d b a b a b b b Mitsou et al, 1988 AS/AT b c c b c b a d b b b a b b b Noyes et al, 1980 AS c b a a c c a a a a a a b b b Otter et al, 1994 AS c a c a c b c d a b b b a b b Rubinstein et al, 1994 AS c b d a b b b d a a b a a c b Sandberg et al, 1986 AS/AT b b e b a b a a a a b b b c b Schwarz et al, 1997 AS c b c a b b a a b b b b b a b Steinbrück et al, 1988 AS c a b c a b b a b b a b b b b Tonino et al, 1986 AS c b c a c b a a a b b b b b b Torg et al, 1976 AT c b c b a c a a 1 b a a b c b Warren et al, 1978 AT c b f b a c a a b b b b b c b Abbreviations: AS, arthroscopy; AT, arthrotomy; MRI, magnetic resonance imaging. * a-f, see criteria list (Appendix) for explanation; 1, only inclusion criteria; 2, only exclusion criteria. Only Lachman test. Only anterior drawer and Lachman test. tive and negative test results, respectively. 4 So, the likelihood ratio incorporates both the sensitivity and specificity of the test and provides a direct estimate of how much a test result will change the odds of having an ACL tear. The DOR, in which the sensitivity and specificity are evaluated as a whole, expresses how much greater the probability of having an ACL tear is for the people with a positive test result than for the people with a negative test result. The DOR is defined as: (true positives true negatives) (false positives false negatives). It is a single summary of diagnostic test performance in which the sensitivity, specificity, LR+, and LR are combined. We chose to present this statistic in this study to enhance comparison among different studies. 16 Subgroup analysis was done to determine whether diagnostic accuracy differed in subgroups defined by the characteristics of the patients. We defined 4 subgroups: patients with acute and chronic ruptures of the ACL and patients to whom the test was applied without and with anesthesia. Regarding the classification of acute and chronic ACL ruptures, we followed the decisions adopted by the authors of the studies included in this meta-analysis. RESULTS Methodological Quality and Study Characteristics The searches in MEDLINE, EMBASE, and CINAHL revealed 2687, 4038, and 743 studies, respectively (325 duplicates were removed, leaving 7143 studies). One article presented in abstract form only was not considered. 76 One study was excluded because the full text of the article could not be obtained. 57 By excluding the rest of the studies considered as not meeting the inclusion criteria, 48 studies remained. After evaluation of the full text of these 48 studies, 21 studies were excluded. Five of the 21 studies did not specify the diagnostic accuracy per test but commented on the composite clinical examination for ligamentous injuries. 36,53,54,58,64 These articles did not include data for specific clinical or surgical examination. Studies that described the diagnostic accuracy of measuring anterior laxity by instruments, such as the KT-1000, were also excluded. 6,7,14,40,70 Three of the 21 excluded studies were review articles. 45,62,67 Eight additional studies did not fulfill the inclusion criteria. 21,25,37,41,43,46,65,66 One study, 55 that was not indexed in 1 of the databases but was identified 270 J Orthop Sports Phys Ther Volume 36 Number 5 May 2006

5 through personal communication, was added to the included studies. Two articles pertained to the same study. 50,52 Overall, we analyzed 28 studies that met our inclusion criteria. 5,8-10,12,13,18,20,26,27,30,32,34,35, 38,39,42,44,48,50,52,55,60,61,63,69,71,72,74 Twenty-seven of these studies were identified in MEDLINE, 10 in EMBASE, and 1 in CINAHL. A detailed description of the methodological quality and study characteristics is presented in Table 2. Two different gold standards, arthroscopy, and arthrotomy, were used in 7 studies, 5,12,27,32,44,48,61 which were each applied to a different group of patients. MRI was used in 2 studies. 8,39 Furthermore, 14 studies applied arthroscopy 9,13,18,26,34,35,38,42,50,51, 55,60,63,69,71 and 5 studies arthrotomy. 10,20,30,72,74 In just 3 studies there was independent, blind comparison of the index test with a reference standard of diagnosis and vice versa. 39,55,69 Additionally, in 3 studies the reference standard was applied regardless of the results of the findings from the index tests (avoidance of verification bias). 9,18,50,52 Only 3 studies mentioned that the index test was applied independently of all other clinical information. 12,55,60 The spectrum of the condition of the experimental group ( diseased patients ) was not specified or unclear in 5 studies. 10,27,39,42,55 ACL ruptures with concomitant lesions to other soft tissue structures of the knee were addressed in 19 studies, 5,8,9,18,20,26,30,32,34,35,38,44, 48,50,52,61,63,71,72,74 whereas 4 studies examined patients with an isolated ACL rupture (small spectrum). 12,13,60,69 The index test was described in sufficient detail to enable its application in one s own TABLE 3A. Anterior drawer test, without anesthesia, whole group. Sensitivity (95% CI) practice in 11 studies. 5,10,12,13,26,27,30,34,50,52,69,72 In 2 studies 55,63 only male patients were included, whereas 9 studies 10,18,27,32,34,39,48,71,72 did not specify the gender of the included patients. The total number of included patients per study ranged from to Furthermore, the average age of the patients ranged from 20 to 33 years. The age of the patients was not specified in 7 studies. 10,18,27,32,34,48,72 Pooled Analysis The pooled sensitivity, specificity, LR+, LR, and DOR for the total population are shown in Tables 3A and 6A and in Tables 4A and 7A for the anterior drawer and the Lachman tests, respectively. The pooled data for the pivot shift test are presented in Tables 5A and 8A. Our analysis shows that the Lachman test is the most accurate test to determine an ACL rupture. Additionally, the pivot shift test shows a very low sensitivity and very high specificity without anesthesia, 24% (95% Confidence Interval [CI], 21-27) and 98% (95% CI, 96-99), respectively. The pooled analysis of the clinical subgroups, patients with acute and chronic lesions, is shown in Tables 3B and 6B for the anterior drawer test. Without anesthesia, the anterior drawer test shows higher sensitivity and specificity in chronic than in acute condition. Tables 4B and 7B and Tables 5B and 8B show the results for the Lachman and the pivot shift tests, respectively. The pivot shift test is very specific both in acute as well as in chronic conditions. Specificity (95% CI) LR+ (95% CI) LR (95% CI) DOR (95% CI) Anderson et al, (14-43) Boeree et al, (42-69) 92 (86-96) 6.7 ( ) 0.5 ( ) 14 (6-31) Bomberg et al, (21-64) 100 (48-100) 5.0 ( ) 0.6 ( ) 8 (0-157) Braunstein, (59-100) 100 (82-100) 33.3 ( ) 0.1 ( ) 259 ( ) DeHaven, (2-23) Donaldson et al, (60-79) Hardaker et al, (11-27) Hughston et al, (37-78) 50 (30-70) 1.2 ( ) 0.8 ( ) 1 (0-4) Jonsson et al, (87-99) Lee et al, (56-93) 100 (94-100) 87.9 ( ) 0.2 ( ) 380 ( ) Liu et al, (43-76) Mitsou et al, (28-54) Mitsou et al, (88-99) Noyes et al, (15-37) 96 (79-100) 5.9 ( ) 0.8 ( ) 8 (1-60) Rubinstein et al, (38-96) 87 (69-96) 5.6 ( ) 0.3 ( ) 20 (3-124) Sandberg et al, (30-48) 97 (88-100) 11.2 ( ) 0.6 ( ) 18 (4-76) Steinbrück et al, (81-98) 91 (87-94) 10.4 ( ) 0.1 ( ) 121 (40-368) Tonino et al, (12-46) 100 (85-100) 12.6 ( ) 0.7 ( ) 17 (1-313) Torg et al, (44-61) 100 (95-100) 82.5 ( ) 0.5 ( ) 171 ( ) Warren et al, (61-80) 77 (56-91) ( ) 0.4 ( ) 8 (3-23) Pooled 55 (52-58)* 92 (90-94)* 7.3 ( )* 0.5 ( )* 21 (8-53)* Sample size aggregate results * P.05. Index test measured independently of all other clinical information. LITERATURE REVIEW J Orthop Sports Phys Ther Volume 36 Number 5 May

6 TABLE 4A. Lachman test, without anesthesia, whole group. Journal of Orthopaedic & Sports Physical Therapy Sensitivity (95% CI) Specificity (95% CI) LR+ (95% CI) LR (95% CI) DOR (95% CI) Anderson et al, (79-98) Boeree et al, (49-75) 90 (84-95) 6.5 ( ) 0.4 ( ) 16 (7-33) Bomberg et al, (65-97) 60 (15-95) 2.2 ( ) 0.2 ( ) 10 (1-83) Cooperman et al, (40-92) 54 (30-77) 1.5 ( ) 0.5 ( ) 3 (1-13) Dahlstedt et al, (82-100) Dahlstedt et al, (85-100) DeHaven, (52-96) Donaldson et al, (95-100) Harilainen, (94-100) 98 (94-99) 40.0 ( ) 0.0 ( ) 1897 ( ) Hardaker et al, (65-82) Jonsson et al, (89-100) Learmonth, (55-79) 94 (89-97) 10.7 ( ) 0.3 ( ) 31 ( ) Lee et al, (72-99) 100 (94-100) ( ) 0.1 ( ) 972 ( ) Liu et al, (82-99) Mitsou et al, (94-100) Rubinstein et al, (60-100) 100 (89-100) 58.2 ( ) 0.1 ( ) 637 ( ) Sandberg et al, (39-57) 97 (88-100) 13.8 ( ) 0.5 ( ) 25 (6-109) Schwarz et al, (80-98) 56 (25-85) 2.1 ( ) 0.1 ( ) 15 (3-72) Steinbrück et al, (74-94) 92 (88-95) 10.7 ( ) 0.1 ( ) 72 (29-180) Tonino et al, (74-98) 100 (85-100) 40.8 ( ) 0.1 ( ) 354 ( ) Torg et al, (92-99) 100 (95-100) ( ) 0.0 ( ) 3754 ( ) Pooled 85 (83-87)* 94 (92-95)* 10.2 ( )* 0.2 ( )* 70 (23-206)* Sample size aggregate results * P.05. Index test measured independently of all other clinical information. TABLE 5A. Pivot shift, without anesthesia, whole group. Sensitivity (95% CI) Specificity (95% CI) LR+ (95% CI) LR (95% CI) DOR (95% CI) Anderson et al, (27-58) Boeree et al, (19-44) 97 (92-99) 8.8 ( ) 0.7 ( ) 12 (4-35) Bomberg et al, (1-29) 100 (48-100) 1.3 ( ) 1.0 ( ) 1 (0-32) Dahlstedt et al, (1-28) Dahlstedt et al, (47-90) DeHaven, (2-23) Donaldson et al, (26-45) Hardaker et al, (20-39) Liu et al, (54-85) Otter et al, (0-71) 82 (57-96) 0.6 ( ) 1.1 ( ) 1 (0-14) Rubinstein et al, (57-100) 89 (73-97) 8.5 ( ) 0.1 ( ) 114 (7-1967) Sandberg et al, (2-11) 100 (94-100) 7.1 ( ) 0.9 ( ) 7 (0-133) Steinbrück et al, (11-35) 99 (97-100) 26.9 ( ) 0.8 ( ) 34 (7-159) Tonino et al, (6-35) 100 (85-100) 8.2 ( ) 0.8 ( ) 10 (1-185) Torg et al, (5-15) 100 (95-100) 14.4 ( ) 0.9 ( ) 16 (1-270) Pooled 24 (21-27)* 98 (96-99)* 8.5 ( ) 0.9 ( )* 12 (5-31) Sample size aggregate results * P.05. P.05. Index test measured independently of all other clinical information. DISCUSSION The purpose of this review was to define the diagnostic accuracy of clinical tests for detecting ruptures of the ACL. The increasing prevalence of ACL injuries coupled with a great financial expense has increased awareness in sports medicine resulting in investigations on how to decrease ACL injury risk. Therefore, it is of socioeconomic importance to correctly diagnose this knee pathology. In this extensive qualitative and quantitative synthesis of the literature, the different values of each test for acute and chronic lesions with or without anesthesia are presented. We included knees with combined lesions, such as ACL ruptures with MCL or lateral collateral ligament 272 J Orthop Sports Phys Ther Volume 36 Number 5 May 2006

7 TABLE 6A. Anterior drawer test, with anesthesia, whole group. Sensitivity (95% CI) Specificity (95% CI) LR+ (95% CI) LR (95% CI) DOR (95% CI) Anderson et al, (67-92) Bomberg et al, (14-55) 60 (15-95) 0.8 ( ) 1.1 ( ) 1 (0-5) DeHaven, (34-69) Donaldson et al, (83-95) Hardaker et al, (40-61) Hughston et al, (86-100) 23 (9-44) 1.3 ( ) 0.1 ( ) 16 (1-293) Jonsson et al, (88-100) Jonsson et al, (91-100) Katz et al, (21-64) 95 (87-99) 8.6 ( ) 0.6 ( ) 14 (3-58) Kim et al, (72-86) Mitsou et al, (84-98) Noyes et al, (42-69) 88 (68-97) 4.5 ( ) 0.5 ( ) 9 (2-33) Sandberg et al, (60-76) 86 (75-94) 5.0 ( ) 0.4 ( ) 13 (6-31) Tonino et al, (37-75) 100 (85-100) 26.0 ( ) 0.4 ( ) 58 (3-1051) Torg et al, (97-100) 100 (95-100) ( ) 0.0 ( ) ( ) Pooled 77 (75-80)* 87 (82-91)* 5.9 ( )* 0.4 ( )* 19 (4-85)* Sample size aggregate results * P.05. Journal of Orthopaedic & Sports Physical Therapy TABLE 7A. Lachman test, with anesthesia, whole group. Sensitivity (95% CI) Specificity (95% CI) LR+ (95% CI) LR (95% CI) DOR (95% CI) Anderson et al, (92-100) Bomberg et al, (71-99) 20 (1-72) 1.1 ( ) 0.5 ( ) 3 (0-35) Dahlstedt et al, (85-100) Dahlstedt et al, (82-100) DeHaven, (78-100) Donaldson et al, (96-100) Hardaker et al, (93-100) Jonsson et al, (88-100) Jonsson et al, (91-100) Katz et al, (60-95) 97 (89-100) 25.8 ( ) 0.2 ( ) 137 (23-811) Kim et al, (95-100) Mitsou et al, (89-100) Sandberg et al, (82-94) 83 (71-91) 5.1 ( ) 0.1 ( ) 38 (16-91) Tonino et al, (78-99) 100 (85-100) 42.3 ( ) 0.1 ( ) 513 ( ) Torg et al, (97-100) 100 (95-100) ( ) 0.0 ( ) ( ) Pooled 97 (96-98)* 93 (89-96)* 12.9 ( )* 0.1 ( )* 131 ( )* Sample size aggregate results * P.05. (LCL) injuries. In addition, we included partial tears in the data analysis to follow the methodology adopted by the authors of the studies included in the meta-analysis. It was not possible to make a distinction between total or partial ruptures of the ACL in some studies. Only in 2 studies 9,44 a definition of a partial ACL tear was reported. Lucie et al 44 defined a partial ACL tear as a ligament which had no loss of competence, whereas Bomberg and McGinty 9 defined it as a tear of a portion of the ACL, leaving the majority of the ligament fibers intact. Based on these definitions we did not consider these tears as ACL ruptures in our data analysis. Furthermore, we included patients who had bilateral ACL injuries and counted these ruptures as 2 independent knees. However, this was just the case in 2 studies 20,50,52 (3 patients). We do not believe that this very small number has any bias effect on the total meta-analysis, nor that this will result in inaccurate results. In addition, some of the studies included only subjects with ACL-deficient knees, or did not report the trueor false-positive test results. It was not possible to calculate specificity (thus LR+, LR, and DOR), in these studies with preselection bias. Lachman Test Our results show that the Lachman test has a good diagnostic accuracy to detect ACL pathology, both in acute as well as in chronic conditions. Recent research from Scholten et al 62 has shown a pooled LITERATURE REVIEW J Orthop Sports Phys Ther Volume 36 Number 5 May

8 TABLE 8A. Pivot shift, with anesthesia, whole group. Sensitivity (95% CI) Specificity (95% CI) LR+ (95% CI) LR (95% CI) DOR (95% CI) Anderson et al, (81-99) Bomberg et al, (36-79) 60 (15-95) 1.5 ( ) 0.7 ( ) 2 (0-16) Dahlstedt et al, (47-87) Dahlstedt et al, (59-96) DeHaven, (45-79) Donaldson et al, (93-100) Hardaker et al, (64-82) Katz et al, (60-95) 98 (92-100) 51.5 ( ) 0.2 ( ) 279 ( ) Kim et al, (84-94) Lucie et al, (83-99) 100 (69-100) 20.7 ( ) 0.1 ( ) 323 ( ) Sandberg et al, (77-91) 100 (94-100) 99.6 ( ) 0.2 ( ) 633 ( ) Tonino et al, (31-69) 100 (85-100) 23.0 ( ) 0.5 ( ) 45 (3-809) Torg et al, (21-37) 100 (95-100) 45.6 ( ) 0.7 ( ) 64 (4-1051) Pooled 74 (71-77)* 99 (96-100)* 20.9 ( )* 0.3 ( )* 75 (11-518)* Sample size aggregate results * P.05. Journal of Orthopaedic & Sports Physical Therapy sensitivity of 86% (95% CI, 76-92) and a specificity of 91% (95% CI, 79-96) for the total population. With a smaller CI, our results are more precise, showing a pooled sensitivity of 85% (95% CI, 83-87) and a specificity of 94% (95% CI, 92-95). Based on 12 studies, we calculated an overall LR+ of 10.2 (95% CI, ) and a LR of 0.2 (95% CI, ) when the test is performed without anesthesia. Recent work from Solomon et al 67 presented a LR+ of 42.0 (95% CI, ) and a LR of 0.1 (95% CI, ) for the Lachman test. However, their data had a wider CI and were based on only 1 study. 39 The position of the knee during this test (20 to 30 of flexion) is a less painful position than the position of the knee during the anterior drawer test; hence, it reduces possible muscle action to protect the knee during testing. Moreover, the secondary restraints do not contribute much to stability in this test position. 49 Rosenberg and Rasmussen 59 measured the tension of the ACL in 20 normal knees and noted that baseline tension in the anteromedial and the posterolateral portions of the ACL was greater at 15 of flexion than at 90. A Lachman test produced maximal tension in the majority of the ACL. This explains the specificity of the Lachman test as an indicator of ACL integrity. Nonetheless, there are some limitations to the test. For example, examiners who have small hands may face difficulties on patients with a large thigh girth. Additionally, the position of the knee is critical because decreasing the flexion angle to 10 may result in decreased excursion of the tibia and a false end point. 20 Pivot Shift Test The specificity of the pivot shift test is very high, namely 98% (95% CI, 96-99). We found, however, very poor sensitivities of 32% (95% CI, 25-38) and 40% (95% CI, 29-52) in acute and chronic conditions, respectively. These data, on which the pooled sensitivity in chronic conditions is based, are very heterogeneous. Interestingly, values between 0% (95% CI, 0-71) 55 and 93% (95% CI, ) 60 were reported. This large range may be due to small sample size, and differences in time from injury, and in the type of clinical setting. It may further be explained by the fact that the prevalence of ACL ruptures in the study of Otter et al 55 was just 15%, and therefore had a smaller chance to diagnose an ACL rupture than in the study of Rubinstein et al, 60 which had a prevalence of 22.5% ACL ruptures. We could not compare our results for the total population with the results from Scholten et al, 62 because they did not calculate pooled sensitivity and specificity due to an insufficient number of studies. Based on 8 studies, we calculated an overall LR+ of 8.5 (95% CI, ) and a LR of 0.9 (95% CI, ) for testing performed without anesthesia. The LR+ and LR of this test were not available in work from Solomon et al 67 and could therefore not be compared to ours. The pivot shift test reproduces the phenomenon of giving way of the knee. A positive test shows anterolateral laxity of the knee, which is most often reported by the patient as a recognizable sensation. The reason for the very low sensitivity in chronic condition 55,69 may be explained by the fact that the patient with a chronic ACL-deficient knee is familiar with this unpleasant phenomenon and will show protective muscle action. Additionally, to perform this test, the MCL must be intact to build up enough contact pressure in the lateral compartment. 20 Moreover, the pivot shift sign is also intimately dependent upon the normal function of the iliotibial tract, which tightens and therefore causes reduction of the displaced tibial plateau at approximately 30 of flexion. 65 Kim et al 35 reported J Orthop Sports Phys Ther Volume 36 Number 5 May 2006

9 TABLE 3B. Anterior drawer test, without anesthesia, acute and chronic subgroups. Sensitivity (95% CI) Specificity (95%CI) LR+ (95%CI) LR (95%CI) DOR (95%CI) Acute Chronic Acute Chronic Acute Chronic Acute Chronic Acute Chronic Anderson et al, (6-42) 60 (27-88) Bomberg et al, (21-64) 100 (48-100) 5.0 ( ) 0.6 ( ) 8 (0-157) DeHaven, (2-23) Donaldson et al, (60-79) Hughston et al, (37-78) 50 (30-70) 1.2 ( ) 0.8 ( ) 1 (0-4) Jonsson et al, (87-99) Liu et al, (43-76) Mitsou et al, (28-54) 95 (88-99) Rubinstein et 76 (38-96) 87 (69-96) 5.6 ( ) 0.3 ( ) 20 (3-124) al, 1994 Steinbrück et al, (81-98) 91 (87-94) 10.4 ( ) 0.1 ( ) 121 (40-368) Pooled 49 (43-55)* 92 (88-95)* 58 (39-76)* 91 (87-94) 1.4 ( ) 8.9 ( ) 0.7 ( ) 0.1 ( ) 2 (1-7) 57 (10-330) Sample size aggregate results * P.05. P.05. Index test measured independently of all other clinical information. LITERATURE REVIEW J Orthop Sports Phys Ther Volume 36 Number 5 May

10 TABLE 4B. Lachman test, without anesthesia, acute and chronic subgroups. Sensitivity (95% CI) Specificity (95%CI) LR+ (95%CI) LR (95%CI) DOR (95%CI) Acute Chronic Acute Chronic Acute Chronic Acute Chronic Acute Chronic Anderson et al, (71-99) 100 (69-100) Bomberg et al, (65-97) 60 (15-95) 2.2 ( ) 0.2 ( ) 10 (1-83) Cooperman et 71 (40-92) 54 (30-77) 1.5 ( ) 0.5 ( ) 3 (1-13) al, 1990 Dahlstedt et al, (85-100) 100 (82-100) DeHaven, (52-96) Donaldson et al, (95-100) Harilainen, (94-100) 98 (94-99) 40.0 ( ) 0.0 ( ) 1897 ( ) Jonsson et al, (89-100) Learmonth, (56-86) 100 (81-100) Liuetal, (82-99) Mitsou et al, (94-100) Rubinstein et 96 (60-100) 100 (89-100) 58.2 ( ) 0.1 ( ) 637 ( ) al, 1994 Steinbrück et al, (73-94) 92 (88-95) 10.7 ( ) 0.1 ( ) 72 (29-180) Pooled 94 (91-96)* 95 (91-97)* 97 (93-99)* 90 (87-94)* 9.4 ( )* 7.1 ( )* 0.1 ( )* 0.2 ( )* 143 ( )* 38 (2-572)* Sample size aggregate results * P.05. Index test measured independently of all other clinical information. 276 J Orthop Sports Phys Ther Volume 36 Number 5 May 2006

11 TABLE 5B. Pivot shift, without anesthesia, acute and chronic subgroups. Sensitivity (95% CI) Specificity (95%CI) LR+ (95%CI) LR (95%CI) DOR (95%CI) Acute Chronic Acute Chronic Acute Chronic Acute Chronic Acute Chronic Bomberg et al, (1-29) 100 (48-100) 1.3 ( ) 1.0 ( ) 1 (0-32) Dahlstedt et al, (1-28) 72 (47-90) DeHaven, (4-17) Donaldson et al, (26-45) Liu et al, (54-85) Otter et al, 0 (0-71) 82 (57-96) 0.6 ( ) 1.1 ( ) 1 (0-14) 1994 Rubinstein et 93 (57-100) 89 (73-97) 8.5 ( ) 0.1 ( ) 114 (7-1967) al, 1994 Steinbrück et al, (11-35) 99 (97-100) 26.9 ( ) 0.8 ( ) 34 (7-159) Pooled 32 (25-38)* 40 (29-52)* 100 (48-100) 97 (95-99)* 1.3 ( ) 7.7 ( ), 1.0 ( ) 0.8 ( )* 1 (0-32) 16 (1-218)* Sample size aggregate results * P.05. P.05. Index test measured independently of all other clinical information. LITERATURE REVIEW J Orthop Sports Phys Ther Volume 36 Number 5 May

12 TABLE 6B. Anterior drawer test, with anesthesia, acute and chronic, subgroups. Sensitivity (95% CI) Specificity (95%CI) LR+ (95%CI) LR (95%CI) DOR (95%CI) Acute Chronic Acute Chronic Acute Chronic Acute Chronic Acute Chronic Anderson et al, (7-43) 61 (27-88) Bomberg et al, (14-55) 60 (15-95) 0.8 ( ) 1.1 ( ) 1 (0-5) DeHaven, (34-69) Donaldson et al, (83-95) Hughston et al, (86-100) 23 (9-44) 1.3 ( ) 0.1 ( ) 16 (1-293) Jonsson et al, (88-100) 98 (91-100) Katzetal, (3-60) 54 (25-81) 97 (89-100) 100 (94-100) 7.0 ( ) 68.6 ( ) 0.8 ( ) 0.5 ( ) 9 (1-72) 147 (7-2869) Kim et al, (95-100) Mitsou et al, (84-98) Pooled 78 (73-82)* 94 (91-97)* 75 (64-83)* 100 (94-100) 1.5 ( ) 68.6 ( ) 0.8 ( ) 0.5 ( ) 4 (1-27) 147 (7-2869) Sample size aggregate results * P.05. P J Orthop Sports Phys Ther Volume 36 Number 5 May 2006

13 TABLE 7B. Lachman test, with anesthesia, acute and chronic subgroups. Sensitivity (95% CI) Specificity (95%CI) LR+ (95%CI) LR (95%CI) DOR (95%CI) Acute Chronic Acute Chronic Acute Chronic Acute Chronic Acute Chronic Anderson et al, (73-99) 100 (69-100) Bomberg et al, (71-99) 20 (1-72) 1.1 ( ) 0.5 ( ) 3 (0-35) Dahlstedt et al, (85-100) 100 (82-100) DeHaven, (78-100) Donaldson et al, (96-100) Jonsson et al, (88-100) 98 (91-100) Katzetal, (40-97) 85 (55-98) 97 (89-100) 100 (94-100) 24.5 ( ) ( ) 0.2 ( ) 0.2 ( ) 107 (13-881) 584 ( ) Kim et al, (95-100) Mitsou et al, (89-100) Pooled 97 (95-99)* 98 (95-99) 91 (82-97)* 100 (94-100) 5.0 ( )* ( ) 0.3 ( ) 0.2 ( ) 18 (0-727)* 584 ( ) Sample size aggregate results * P.05. P.05. LITERATURE REVIEW J Orthop Sports Phys Ther Volume 36 Number 5 May

14 TABLE 8B. Pivot shift, with anesthesia, acute and chronic subgroups. Sensitivity (95% CI) Specificity (95%CI) LR+ (95%CI) LR (95%CI) DOR (95%CI) Acute Chronic Acute Chronic Acute Chronic Acute Chronic Acute Chronic Bomberg et al, (36-79) 60 (15-95) 1.5 ( ) 0.7 ( ) 2 (0-16) Dahlstedt et al, (47-99) 83 (59-96) DeHaven (45-79) Donaldson et al, (93-100) Katz et al, (52-100) 85 (55-98) 100 (94-100) 98 (92-100) ( ) 53.3 ( ) 0.2 ( ) 0.2 ( ) 720 ( ) 341 ( ) Kim et al, (84-94) Lucie et al, (83-99) 100 (69-100) 20.7 ( ) 0.1 ( ) 323 ( ) Pooled 85 (80-90)* 89 (83-93) 97 (91-100)* 98 (92-100) 12.6 ( )* 53.3 ( ) 0.2 ( )* 0.2 ( ) 67 (1-3604)* 341 ( ) Sample size aggregate results * P.05. P J Orthop Sports Phys Ther Volume 36 Number 5 May 2006

15 false-negative results of the pivot shift test in chronic condition, caused by reattachment of a torn ACL to the proximal portion of the PCL in 7 of the 15 patients, as confirmed by arthroscopic examination. It is commonly believed that the accuracy of a complex test maneuver such as the pivot shift test may increase with experience. 67 Anterior Drawer Test The anterior drawer test is widely used in the diagnosis of ACL injuries. Our results show that the test has unacceptable low sensitivity and specificity for use in the clinical setting, especially in acute condition (49% [95% CI, 43-55] and 58% [95% CI, 39-76], respectively). Torg et al 72 noted 3 possible causes for a false-negative anterior drawer test in acute condition, especially in isolated ACL tears. First, the hemarthrosis and reactive synovitis may preclude knee flexion to 90, hindering the proper performance of the test. Second, protective muscle action of the hamstrings secondary to joint pain provides a vector force opposite to the anterior translation of the tibia. Third, the posterior horn of the medial meniscus becomes buttressed against the posteriormost margin of the medial femoral condyle and may preclude anterior translation of the tibia. 72 In addition, this test is strongly affected by concomitant injury and becomes increasingly positive if secondary restraints are injured. In chronic condition, when there is less effusion, pain, muscular protection, and probably progressive knee laxity due to failure of the secondary restraints, the sensitivity and specificity of the anterior drawer test are much better (92% [95% CI, 88-95] and 91% [95% CI, 87-94], respectively). For the total population, we calculated a pooled sensitivity and specificity of 55% (95% CI, 52-58) and 92% (95% CI, 90-94), respectively. With a somewhat wider CI, Scholten et al 62 presented almost the same values of the pooled sensitivity and specificity. They calculated a sensitivity of 62% (95% CI, 42-78) and a specificity of 88% (95% CI, 83-92). The anterior drawer test was found to have the best diagnostic value for the LR+, LR, and DOR in chronic lesions as compared to the other tests that were analyzed in this study. Based on 12 studies, we calculated an overall LR+ of 7.3 (95% CI, ) and a LR of 0.5 (95% CI, ) when tested without anesthesia. Recent work from Solomon et al 67 indicated a LR+ of 3.8 (95% CI, ) and a LR of 0.3 (95% CI, ). However, these data, with a wide CI, were based on 3 studies only. 10,30,39 The high diagnostic values in chronic condition could be attributed to the fact that patients who initially had isolated ACL injury may develop progressive knee laxity over time due to the failure of the secondary restraints. Posttest Probabilities in Acute and Chronic Conditions Sensitivity and specificity only describe how abnormality in this case a rupture of the ACL predicts a test result. 17 However, for clinical practice it is essential to know how a test result predicts abnormality. Likelihood ratios provide this information. According to a prospective investigation from Noyes et al, % of patients with acute knee injuries (defined as individuals who suffered a traumatic knee injury associated with a rapid onset of swelling [ie, a hemarthrosis] with or without the sensation of a pop) have a completely torn ACL. In other words, the pretest probability of having an ACL rupture in these patients is 44.0%. We calculated that in acute cases the LR+ of the Lachman test is 9.4 (95% CI, ). From this value the posttest probability can be calculated, which is 88.1%. So a positive Lachman test increases the probability of a ruptured ACL from 44.0% to 88.1% in this population. A LR of 0.1 (95% CI, ), in case of a negative Lachman test, changes the pretest probability from 44.0% of having an ACL rupture to a posttest probability of 7.3% in patients with acute knee injuries, as described by Noyes et al. 50 A negative Lachman test therefore decreases the probability of having a ruptured ACL from 44.0% to 7.3%. This is shown in the nomogram in Figure 2. According to an other prospective investigation from Steinbrück et al, % of patients with chronic knee injuries (not defined in this study) have a completely torn ACL. So, the pretest probability of having an ACL rupture in these patients is 22.3%. We calculated that in chronic cases the LR+ of the Lachman test is 7.1 (95% CI, ). From this value the posttest probability of having a ruptured ACL can be calculated, which is 67.1%. So a positive Lachman test in chronic conditions increases the probability of a ruptured ACL from 22.3% to 67.1%. A LR of 0.2 (95% CI, ) results in a posttest probability of 5.4% in case of a negative Lachman test. A negative Lachman test, therefore, decreases the probability of having a ruptured ACL from 22.3% to 5.4%. The LR+ of the pivot shift test in acute condition is 1.3 (95% CI, ). With a pretest probability of an ACL rupture of 44.0%, 50 this results in a posttest probability of a torn ACL of 50.5%, in case of a positive pivot shift test. A LR of 1.0 (95% CI, ) in case of a negative pivot shift test does not change the pretest probability of 44.0% of having a ruptured ACL, the posttest probability is also 44.0%. So, only a positive pivot shift test gives a relatively small increase of 6.5% of the probability for having a ruptured ACL. The LR+ of the pivot shift test in chronic condition is 7.7 (95% CI, ), with a pretest probability of LITERATURE REVIEW J Orthop Sports Phys Ther Volume 36 Number 5 May

16 anterior drawer test decreases the probability of having a ruptured ACL by 8.5%. The anterior drawer test shows a LR+ of 8.9 (95% CI, ) in chronic conditions. Assuming that the pretest probability of having an ACL rupture is 22.3%, 69 this results in a posttest probability of a torn ACL of 71.9%. A positive anterior drawer test in chronic conditions therefore increases the probability of having an ACL rupture from 22.3% to 71.9%. A LR of 0.1 (95% CI, ) results in a posttest probability of 2.8% of having an ACL rupture in case of a negative anterior drawer test. Therefore, a negative anterior drawer test decreases the probability of having an ACL rupture by 19.5%. Journal of Orthopaedic & Sports Physical Therapy FIGURE 2. Nomogram: posttest probability using the Lachman test to detect an anterior cruciate ligament (ACL) rupture. A: With a LR+ of 9.4 (95% CI, ), the pretest probability of 44.0% of having an ACL rupture changes after a positive Lachman test to a posttest probability of 88.1% of having a ruptured ACL. B: A LR of 0.1 (95% CI, ) in case of a negative Lachman test decreases the pretest probability of 44.0% of having an ACL rupture to a posttest probability of 7.3% of having a ruptured ACL. having a torn ACL of 22.3%. 69 This results in a posttest probability of an ACL rupture of 68.8% in case of a positive test. So a positive pivot shift test in chronic conditions increases the probability of a ruptured ACL from 22.3% to 68.8%. A LR of 0.8 (95% CI, ) results in a posttest probability of a ruptured ACL of 18.7% in case of a negative pivot shift test. A negative pivot shift test therefore results in a small decrease of the probability of having an ACL rupture from 22.3% to 18.7%. The anterior drawer test shows a LR+ of 1.4 (95% CI, ) in acute conditions. Again assuming that the pretest probability of having an ACL rupture is 44.0%, 50 this results in a posttest probability of a torn ACL of 52.4%. A positive anterior drawer test, therefore, increases the probability of having an ACL rupture by 8.4%. A LR of 0.7 (95% CI, ) in case of a negative anterior drawer test changes the pretest probability from 44.0% of having a ruptured ACL to a posttest probability of 35.5%. So, a negative The Role of Anesthesia in the Diagnostic Accuracy of Testing The accuracy of the tests may have been affected by factors such as swelling, pain, and protective muscle action. Anesthesia eliminates protective muscle guarding, which enhances the accuracy in confirmatory tests such as the pivot shift. So, anesthesia adds to the diagnostic precision. The overall sensitivity of the pivot shift test for example is 74% (95% CI, 71-77) under anesthesia, but in the alert patient this percentage decreases to 24% (95% CI, 21-27). Such change was not present when analyzing the overall specificity of this test. There were a few studies 18,30,63,72,73 in which the anterior drawer or the pivot shift tests could not be performed or were inconclusive without anesthesia, because of pain or muscular protection. In these cases, the authors reported the number of knees or a percentage of the total participating knees in which the test was equivocal. In our data analysis we regarded these cases as false or as true negative. Limitations of the Study Valid estimates of the diagnostic accuracy of the 3 tests for detecting ACL ruptures are generalizable only to settings that have a similar spectrum of patients. Furthermore, technical details of the tests may also vary from one setting to another and limit generalizability. 31 Other limitations must be addressed in this study. First, we made the methodological choice to include partial ACL tears in the data analysis. Second, this meta-analysis is based on heterogeneous data. This heterogeneity might be caused by differences in study design, setting, the spectrum of the diseased and nondiseased groups, the threshold of the index test, and the reference standard. 15 The number of patients examined in the studies reviewed was an important determinant of the precision of the estimate of the accuracy of the examination. This is reflected in the confidence interval, which is smaller 282 J Orthop Sports Phys Ther Volume 36 Number 5 May 2006