An examination of the selective tissue tension scheme, with evidence for the concept of a capsular pattern of the knee.

Key Words: Clinical decision making, Knee, Tests and measurements.

The selective tissue tension scheme of James Cyriax[1] is an evaluation method commonly used by physical therapists. Cyriax's scheme consists of active-range-of-motion (AROM), passive-range-of-motion (PROM), and resistive tests, followed by palpation of anatomical structures. According to Cyriax, active movements indicate the patient's willingness to move, the available AROM, and the muscular power available. Resistive movements can be used, according to Cyriax, to assess the status of the contractile structures (muscle, tendon) around the joint,[1](pp70-71) Passive movements are supposed to test noncontractile structures. In Cyriax's scheme, palpation is performed to detect any deformity or inflammatory signs, including warmth or swelling.[1](p69)

In the assessment of PROM, Cyriax contended that the examiner should assess the available PROM, the nature of the end-feel for the motion, and the relationship of the onset of pain with the onset of resistance during PROM (pain-resistance sequence [PRS]). The PRS purportedly reflects the acuity of the inflammatory process. Cyriax contended that pain occurring prior to resistance to movement indicates an acutely inflamed joint, that pain that is synchronous with resistance indicates a less acutely inflamed joint, and that pain occurring after resistance indicates a noninflamed joint.[1](p77) Cyriax proposed that, by evaluating the PRS, the clinician can judge the acuity of the patient's condition and can determine how aggressively to proceed with treatment. For example, if manual stretching of a joint is to be used to regain lost motion, an aggressive approach would be indicated with a judgment of "pain after resistance," according to Cyriax's concept of the PRS.

Cyriax further recommended assessing passive motion for each movement of a joint in order to discern patterns of motion restrictions. A "capsular pattern" is a proportional motion restriction unique to each joint that indicates irritation of the entire synovial membrane or joint capsule, as occurs with an active inflammatory process (arthritis) or degenerative joint changes (arthrosis).[1](p77) According to Cyriax, motion restrictions in proportions other than the capsular pattern are supposed to occur in lesions that are capable of restricting motion, but that are localized in such a way that the whole joint is not involved. Cyriax stated that "noncapsular patterns" fall into 1 of 3 categories: ligamentous adhesions (eg, posttraumatic medial collateral ligaments-adhesion at the knee), internal derangements (eg, meniscal tear), and extra-articular lesions (eg, bursitis, muscle injury).[1](pp80-82) He did not, however, supply any evidence to support this contention. Cyriax stated that the purposes of the selective tissue tension scheme are (1) to clarify the acuity (or severity) of the injury[1](p77) and (2) to identify the structure most responsible for the patient's pain.[1](p73) The degree of severity is based largely on the PRS. According to Cyriax, in order to identity the structure involved, the examiner should consider whether a contractile structure or a noncontractile structure is involved.[1](p63) If a noncontractile structure is thought to be involved, the examiner is supposed to judge whether the pain arises from a generalized involvement of the entire joint (ie, the joint capsule or synovium) or from a more localized pathology not involving the entire joint (eg, a ligamentous adhesion).[1](pp74-75) Cyriax proposed that the later decision is aided by the particular pattern of passive motion restriction found on examination. According to Cyriax, a capsular pattern of restriction is more indicative of involvement of the whole joint (eg, arthritis or arthrosis), and a noncapsular pattern of restriction indicates involvement of specific structures around the joint, such as soft tissue contracture or internal derangement.

For example, Cyriax defined the capsular pattern of the knee as "great limitation of flexion and slight limitation of extension."[1](p600) He did not, however, operationally define "great limitation" or "slight limitation." According to Cyriax, a patient with a substantial loss of flexion and no loss of extension during PROM of the knee would have a noncapsular pattern, and the limitation would more likely be caused by a contracture of the knee's extensor mechanism or an internal derangement than by involvement of the whole joint (ie, capsule, synovium). A patient with a capsular pattern of the knee (gross loss of flexion, with a slight limitation of extension) would have pathology more likely involving the joint capsule or synovium. According to Cyriax, joint mobilization would be indicated for the patient with a capsular pattern, with aggressiveness dictated by the acuity of the inflammatory status, as determined by the PRS. For the patient with a noncapsular pattern, treatment is supposed to be directed toward the pathology, and techniques such as cross-friction massage or stretching of the contracted extensor mechanism would be indicated. Cyriax supplied no data to support the effectiveness of these techniques.

There are 2 recent reports of attempts to evaluate Cyriax's selective tissue tension scheme, and the results vary. Pellecchia et al[2] reported on the interrater reliability of examiners' classifications of patients with shoulder pain. Active range of motion, PROM, and resistive range of motion (ROM) were evaluated, and the therapists placed the patients into a diagnostic category based on the results. The authors reported agreement between the 2 therapists regarding the diagnostic classification of 19 of the 21 patients tested, and they concluded that the Cyriax evaluation scheme was highly reliable in the assessment of shoulder pain.[2] Hayes et al[3] evaluated the passive component of Cyriax's scheme in patients with osteoarthritis of the knee. The authors concluded the validity of the passive components for identifying patients with osteoarthritis of the knee was questionable.[3]

The primary purpose of this study was to examine the relationship between the ratio of PROM restrictions of the knee and the expectation of a capsular pattern based on Cyriax's definitions and, therefore, to examine one of the premises of Cyriax's classification scheme. According to Cyriax, a capsular pattern (ie, a great limitation of flexion, with a slight limitation of extension[1](p600)), is expected in patients with either an acutely inflamed joint based on the presence of classic inflammatory signs (arthritis) or a diagnosis of degenerative changes of the joint (arthrosis). We explored the ratio of loss of extension to loss of flexion during PROM in patients with and without evidence of arthritis or arthrosis. We hypothesized that patients with evidence of arthritis or arthrosis will tend to demonstrate a capsular pattern of PROM restriction (ie, a great loss of flexion, with a slight loss of extension). Based on an examination of the data, we selected a definition of a capsular pattern that maximizes the discrimination between patients with and without evidence of arthritis and arthrosis. After this definition was determined from the data, we performed statistical tests to determine whether there was an association between arthritis or arthrosis and the presence of a capsular pattern of PROM restriction.

A secondary purpose of the study was to evaluate the relationship between the inflammatory status of the joint and the PRS, as well as the chronicity (ie, time from injury or surgery) of the subjects' condition and the PRS. We hypothesized that the PRS would be associated with the inflammatory status of the joint, but would not be associated with the chronicity of the subjects' condition.

Method

Inclusion Criteria

Subjects eligible for this study were individuals referred to physical therapy centers for treatment of unilateral knee dysfunction. The subjects were questioned regarding prior injuries to the contralateral knee. Any patient reporting previous injuries or surgeries involving the contralateral knee were excluded. No patient was excluded based on diagnosis, chronicity, or surgical status of the involved knee. The diagnosis made by the referring physician was noted by the subjects' physical therapist. All subjects were questioned by their physical therapist whether radiographs of the knee were taken. Ail data were collected during the subjects' initial evaluation prior to beginning physical therapy.

Procedure

Data for this study were collected in 15 centers by 33 participating therapists. Each therapist was provided with instructions for performing the measurement of PROM and tests of inflammation. Operational definitions of the PRS were provided, based on the descriptions given by Cyriax.[1](p77) No additional training was provided to the therapists. The subjects' diagnosis, involved knee, surgical history, history of present injury, the patient-reported results of any diagnostic imaging studies performed, and date of onset or the date of surgery were recorded.

Range-of-motion measurements. Extension PROM of the knee joint was measured with the subject positioned supine. The heel was elevated on a bolster to allow for full hyperextension, if present. Flexion PROM was measured with the subject positioned supine and the hip initially in extension. Measurements of PROM for flexion and extension were recorded for each knee. Difference scores for flexion and extension were calculated by subtracting the measurement of the uninvolved knee from the measurement of the involved knee. The ratio of extension loss to flexion loss was calculated by dividing the difference score for extension by the difference score for flexion. Subjects with a difference score for flexion of zero were considered to have a ratio of zero to avoid undefined values in the data analysis. Subjects with a difference score for extension of zero would also have a ratio of zero. If the involved knee had greater motion than the uninvolved knee in either flexion or extension, the ratio of the difference scores had a negative value.

Assessment of the pain-resistance sequence. The PRS was assessed during the measurement of flexion PROM. The examiner first asked each subject to rate his or her baseline level of pain from 0 to 10, with 0 representing no pain and 10 representing the worst pain imaginable, with the knee relaxed in an extended position. The examiner then moved the knee passively into flexion, and the subject was asked when the pain increased above the baseline level during this motion. If the limitation of PROM for flexion was encountered before the subject reported an increase in pain, mild pressure was applied over the subject's anterior tibia to move the knee farther into flexion. The subject was again asked whether the pain had increased above the baseline level. The examiner recorded whether the point of increased pain occurred before, during, or after the limitation of passive motion was encountered.

Assessment of inflammatory status. The cardinal signs of inflammation are pain, redness, warmth, and swelling. All subjects in this study had some degree of pain; therefore, other methods were used to assess for the presence of the 3 inflammatory signs other than pain.[4] These methods were (1) visual inspection for redness, (2) palpatory assessment for warmth, (3) a patellar tap test, and (4) a fluctuation test. Each of these tests was judged by the examiner as either positive or negative for the presence of signs of inflammation.

The examiner first visually examined the involved knee for the presence of redness as compared with the uninvolved knee. The examiner then palpated the anterior aspect of the involved knee for the presence of increased temperature as compared with the uninvolved knee. The patellar tap and the fluctuation test were described by Cyriax as tests for the presence of swelling in the knee joint.[1](p597) The patellar tap is performed with the subject positioned supine. The examiner presses on the suprapatellar pouch, then taps on the patella. If swelling is present, the patella will, in theory, be lifted off the femur and can be tapped down onto the femur. If swelling is not present, the patella should remain in contact with the femur. The fluctuation test is also performed with the subject positioned supine. The examiner places the thumb and finger of one hand around the patella. The other hand is used to push any fluid from the suprapatellar pouch. If swelling is present, the finger and thumb should be pushed apart. If swelling is not present, no movement is supposed occur. We anticipated that the reliability of judgments made with these tests individually could be questionable. We therefore considered the joint to be inflamed when 2 or more of the tests were judged to be positive for signs of inflammation. This was done to avoid the classification of a joint as inflamed based on a single, potentially unreliable judgment.

Categorization of the joint. An inflamed joint was considered to have "arthritis," as the term was used by Cyriax.[1](p77) Subjects with 2 or more tests that were positive for signs of inflammation, therefore, were defined as having arthritis. Subjects whose referring physician diagnosed them as having unilateral knee osteoarthritis, as confirmed by radiographs, were defined as having arthrosis. These 2 groups of subjects (subjects with arthritis and subjects with arthrosis) should have a capsular pattern, according to Cyriax.[1](p77) All other subjects (subjects without evidence of arthritis or arthrosis) would not be expected, according to Cyriax, to have a capsular pattern.

Interrater reliability. Reliability is a precursor to validity. Interrater reliability of the 4 inflammation tests, the categorization of a subject's knee as inflamed or noninflamed, the PRS, the measurements of knee flexion and extension during PROM, and the ratio of extension loss to flexion loss, therefore, was determined on a subset of 35 subjects. These subjects were examined by their treating therapist and then examined by one of the researchers (JMF) during the same session, without any treatment or further evaluation in the intervening period. The results of the first examination were not available to the second therapist. Eight different therapists at 2 centers participated in the collection of reliability data. No additional instructions or training were provided to these therapists.

Interrater reliability for the 4 inflammation tests (warmth, redness, patellar tap, and fluctuation), the PRS, and the categorization of the involved knee as inflamed or not inflamed was analyzed with Cohen's kappa coefficients.[5] Interrater reliability of the PRS measurements was analyzed using a weighted kappa with symmetrical kappa weights.[6] Interrater reliability for measurements of PROM and for the ratio of extension loss to flexion loss was analyzed using intraclass correlation coefficients (ICC[2,1]).[7]

Interrater reliability values for categorical variables are given in Table 1. Kappa coefficients for the inflammatory tests ranged from .21 to .66. Categorization of the joint as inflamed or noninflamed based on the presence of 2 or more inflammatory signs showed substantial clinical agreement ([Kappa]=.76).[8] Judgments of the PRS showed a weighted kappa of .28. Interrater reliability values and standard errors of measurement for the PROM measurements are presented in Table 2. The mean ratio of PROM loss for these 35 subjects was 0.3 (SEM=0.16, ICC=.85). Intraclass correlation coefficients for PROM measures ranged from .72 to .97.

Table 1.
Interrater Reliability for 35 Subjects for Judgments on
Categorical Variables

                                                      Percentage of
Variable                               Kappa            Agreement

Fluctuation test                       .37(a)             .71
Patellar tap test                      .21(a)             .71
Palpation for warmth                   .66(a)             .83
Visual inspection for redness          .21(a)             .85
Categorization of the
inflammatory status of the joint       .76(a)             .89
Pain-resistance sequence               .28(b)             .74

(a) Coefficient represents use of the kappa statistic.(5)

(b) Coefficient represents use of the weighted kappa statistic.(6)

Table 2.
Reliability Coefficients for 35 Subjects for Continuous Variables

Variable                                  [bar] X           SD

Flexion (uninvolved side)              138[degrees]    6.8[degrees]
Flexion (involved side)                118[degrees]   20.6[degrees]
Extension (uninvolved side)              4[degrees]    3.1[degrees]
Extension (involved side)                0[degrees]    6.9[degrees]
Ratio (extension loss/flexion loss)      0.3               0.4

Variable                                         Range

Flexion (uninvolved side)                95[degrees]-155[degrees]
Flexion (involved side)                  46[degrees]-151[degrees]
Extension (uninvolved side)              -9[degrees]-15[degrees]
Extension (involved side)               -34[degrees]-15[degrees]
Ratio (extension loss/flexion loss)              -3.0-6.6

Variable                                 ICC(b)         SEM(b)

Flexion (uninvolved side)                  .80       3.0[degrees]
Flexion (involved side)                    .97       3.9[degrees]
Extension (uninvolved side)                .72       1.7[degrees]
Extension (involved side)                  .94       1.7[degrees]
Ratio (extension loss/flexion loss)        .85       0.16

(a) Intraclass correlation coefficient calculated using equation (2,1) from Shrout and Fleiss.[7]

(b) Standard error of measurement calculated as SD (1 - ICC) 1/2.

Data Analysis

We followed a step-by-step process outlined by Sackett et al[9] for interpreting clinical test results.

Histogram construction. The first step was construction of a histogram showing the number of subjects with and without the target disorder (arthritis or arthrosis), given a certain value of the test result (ratio of extension loss to flexion loss). The upper and lower limits of the ratio of extension loss to flexion loss were determined by examining the histogram and selecting cutoff points that appeared to maximize the differentiation of subjects categorized as with or without arthritis or arthrosis of the knee.

Receiver operating characteristic curve construction. After selecting the cutoff points, subjects were divided into 4 mutually exclusive groups:

1. True positive: subjects with evidence of arthritis or arthrosis with a capsular pattern.

2. False positive: subjects without evidence of arthritis or arthrosis with a capsular pattern.

3. False negative: subjects with evidence of arthritis or arthrosis with a noncapsular pattern.

4. True negative: subjects without evidence of arthritis or arthrosis with a noncapsular pattern.

The number of subjects in each group was displayed in a contingency table (Fig. 1), allowing for the calculation of sensitivity and specificity values. Sensitivity (true-positive rate) describes the test's ability to detect the target disorder when present. Specificity (true-negative rate) describes the test's ability to identify the absence of the target disorder when not present.[9]

Figure 1. Example of general format of a 2 X 2 table for the description of the diagnostic value of a test result, with formulas for the calculation of sensitivity, specificity, and predictive values.

                       Target Disorder         Target Disorder
                           Present                   Absent
Positive Test Result   A True Positives        C False Negatives
Negative Test Result   B True Positives        D True Negatives
                       Sensitivity (A/[A+B])   Specificity (D/[B+D])

The sensitivity and specificity values for given levels of the test result are used to validate the choice of cutoff points by calculating sensitivity and specificity values for different levels of the test result and by graphing the pairs of values as a receiver operating characteristic (ROC) curve.[10] An ROC curve has the sensitivity values expressed as a proportion along the Y axis and the 1 - specificity values expressed as a proportion on the X axis. A perfect test (100% specific and 100% sensitive) would be located in the upper left-hand corner of the graph. The point on the ROC curve closest to the upper left-hand corner, therefore, is considered the "best" cutoff point for the test result.[9] We graphed the sensitivity and specificity values as a ROC curve for the ratios of ROM loss at interval widths of 0.10, and we used the ratio closest to the upper left-hand corner as the cutoff point for defining a capsular pattern.

Contingency table construction. Using this cutoff point, a contingency table was constructed, sensitivity and specificity values were calculated, and a positive likelihood ratio (sensitivity/1-specificity) was determined. Likelihood ratios are becoming increasingly popular for characterizing the value of test results.[11-13] The positive likelihood ratio describes the odds that a given test result would be expected in a subject with (as opposed to without) the target disorder.[9] A positive likelihood ratio of 1 indicates a test that is of no value because it does not change the odds of finding the target disorder. Ratios of greater than 1 indicate a test that increases the likelihood of correctly classifying a subject based on the test result, and ratios of less than 1 indicate a test in which more subjects will be classified incorrectly after the test result is known.[9,11,14] Confidence intervals were determined for the sensitivity, specificity, and likelihood ratio according to the method of Simel et al.[12]

Chi-square tests of association. Chi-square tests of association were used to test the hypotheses concerning the association between (1) subjects with arthrosis or arthritis and the presence of a capsular pattern of PROM restriction, (2) the inflammatory status of the joint and the PRS, and (3) the chronicity of the condition and the PRS. The value of a chi-square test statistic can be greatly influenced by large sample sizes.[15] We therefore set the significance level at .01 for each of the 3 chi-square tests. Cramer contingency coefficients ([V.sup.2]=[chi square]/[N(q-1)], where q is either the number of rows or the number of columns, whichever is smaller) were calculated for each chi-square test as a measure of the degree of association between row and column data in each contingency table.[16] Chi-square analyzes were conducted using the SPSS statistical package.(*)

Results

Data were collected on 152 subjects. Table 3 shows the descriptive subject data. The average age of the sample was 40.0 years (SD = 15.9, range = 13-82). The right side was involved in 71 subjects (46.7%), and the left side was involved in 81 subjects (53.3%). Seventy-seven subjects (50.7%) had undergone knee surgery, and 75 subjects (49.3%) had not undergone knee surgery. Chronicity was divided into 3 categories: acute ([is less than] 2 weeks from onset of pain or surgery), subacute (2-6 weeks from onset of pain or surgery), and chronic ([is greater than] 6 weeks from onset of pain or surgery). Forty-five subjects (29.6%) were classified as acute, 56 subjects (36.8%) were classified as subacute, and 51 subjects (33.6%) were classified as chronic.

Table 3.
Descriptive Data of Subjects(a)

                             Total       Capsular    Noncapsular
                             Sample      Pattern      Pattern
                            (N = 152)    (n = 76)     (n =76)

Age (y)
  [bar] X                      40.0        41.8         38.0
  SD                           15.9        16.6         14.5
  Range                       13-82        21-82       13-77
Involved side
  Right                         71          37           34
  Left                          81          39           42

Diagnosis
  Arthroscopic
   surgery                      41          20           21
  ACL reconstruction            17          15           2
  Patellofemoral
   dysfunction                  15           5           10
  Osteoarthritis                12          10           2
  ACL deficiency                12           4           8
  Internal
   derangement                  11           4           7
  Other (nonsurgical)           25           6           19
  Other (postsurgical)          19          12           7
Pain-resistance
   sequence
  Before                        56          31           25
  During                        71          39           32
  After                         25           6           19
Chronicity
  Acute (<2 wk)                 45          30           15
  Subacute (2-6 wk)             56          20           36
  Chronic (>6 wk)               51          26           25
Inflammatory test
  Warmth                        82          50           32
  Redness                       27          21           6
  Patellar tap                  51          35           16
  Fluctuation                   85          57           28
Categorization of
   joint status
  Inflamed                      71          59           12
  Noninflamed                   81          17           64

(a) Pain-resistance sequence is based on the measurement of flexion range of motion. A capsular pattern was defined as a ratio of extension loss to flexion loss between the values of 0.03 and 0.50. Categorization of joint status is based on the presence of 2 or more inflammatory tests. ACL=anterior cruciate ligament.

Descriptive statistics for the 4 inflammatory tests and the categorization of the inflammatory status of the involved knee are also given in Table 3. Based on the definition of inflammatory status used in this study (2 or more positive tests), 71 knees (46.7%) were considered inflamed, and 81 knees (53.3%) were considered noninflamed.

A capsular pattern of restriction was expected for subjects categorized as having inflamed knees (arthritis) or diagnosed with arthrosis. Seventy-one subjects were categorized as having an inflamed joint, and 12 subjects were diagnosed with arthrosis by radiographs. Four of these 12 subjects were also categorized as having an inflamed joint. Thus, 79 subjects were categorized as having arthritis or arthrosis, and 73 subjects were categorized as not having arthritis or arthrosis.

Histogram Construction

The histogram constructed showing the number of subjects with and without arthritis or arthrosis, given different levels of the ratio of PROM loss, is presented in Figure 2. According to Cyriax's definition of a capsular pattern, subjects with a negative ratio of extension loss to flexion loss, indicating greater motion of the involved knee, or a ratio of zero, indicating equal motion in flexion or extension, would be considered to have a noncapsular pattern and would not be expected to have arthritis or arthrosis. Negative ratios or a ratio of zero, therefore, had to fall outside the range of ratios of PROM loss defining a capsular pattern. It was clear from the histogram that small positive ratios were associated with a greater likelihood of arthritis or arthrosis. Small positive ratios indicate a much higher loss of flexion than extension and, therefore, are consistent with Cyriax's definition. The smallest positive ratio found in any subject was 0.03. We therefore selected a ratio of 0.03 as the lower limit defining a capsular pattern. The choice of an upper limit was not as clear from the histogram. A ratio of 0.50 (flexion loss twice as great as extension loss) was selected. From the histogram, it appeared that defining a capsular pattern as a ratio of extension loss to flexion loss between 0.03 and 0.50 maximized discrimination between subjects with and without evidence of arthritis or arthrosis. We believe that these limits are also consistent with Cyriax's definition of a capsular pattern (great limitation of flexion, with a slight limitation of extension).

[Figure 2 ILLUSTRATION OMITTED]

Receiver Operating Characteristic Curve Construction

The ROC curve constructed from the sensitivity and specificity values for each 0.10 interval of the ratio of PROM loss is shown in Figure 3. The lower bound was maintained at 0.03 because this value clearly maximized the discrimination between knees with and without arthritis or arthrosis. This lower bound, in our opinion, also was consistent with Cyriax's definition of a capsular pattern because ratios of zero indicate no loss of either flexion or extension during PROM and negative ratio values indicate an excess of flexion or extension during PROM on the involved side. The ratio closest to the upper left-hand corner of the graph coincided with the upper limit value (0.50) determined from the histogram. The capsular pattern of the knee, therefore, was defined as a ratio of extension loss to flexion loss between 0.03 and 0.50. Descriptive statistics of the subjects with a capsular or noncapsular pattern of PROM restriction are given in Table 3.

[Figure 3 ILLUSTRATION OMITTED]

Contingency Table Construction

The contingency table constructed using this definition of a capsular pattern (ie, 0.03-0.50) is shown in Table 4. Sensitivity was calculated as 74.7%, and specificity was calculated as 76.7%, with a likelihood ratio of 3.20. Based on these results, a subject with a capsular pattern was 3.2 times more likely than not to have evidence of arthritis or arthrosis of the knee.

Table 4.
Relationship Between the Presence of Arthritis or Arthrosis and a
Capsular Pattern of Restriction of Range of Motion(a)

                                   Arthritis or       Arthritis or
                                    Arthrosis         Arthrosis
                                     Present          Not Present

Capsular pattern present                59               17
Noncapsular pattern present             20               56

Sensitivity=59/79 X 100%=74.7% (67.8, 81.6)

Specificity=56/73 X 100%=76.7% (70.0, 83.4)

Likelihood ratio=.747/(1-.767)=3.20 (2.00, 5.12)

[chi square] =40.09

P < .000001

[V.sup.2] = .264

(a) A capsular pattern was defined as a ratio of extension loss to flexion loss between the values of 0.03 and 0.50. Values in parentheses represent the 95% confidence interval.(11) Significance set at P < .01. [V.sup.2] = Cramer contingency coefficient.

Chi-square Tests of Association

The hypothesis of an association between arthritis or arthrosis and the presence of a capsular pattern was confirmed ([chi square]=40.09, P [is less than] .000001, [V.sup.2]=.264). The pattern of PROM restriction explained 26.4% of the variability in the presence or absence of arthritis or arthrosis (Tab. 4).

The associations between the PRS and chronicity and between the PRS and inflammatory status are shown in Table 5. The hypothesis of no association between the PRS and chronicity was rejected ([chi square]=16.10, P=.0029, [V.sup.2]=.053). The hypothesis of an association between the PRS and the inflammatory status of the joint was confirmed ([chi square]= 18.74, P [is less than] .00001, [V.sup.2]=.123). The chronicity and inflammatory status explained .5.3% and 12.3% of the variability in the PRS, respectively.

Table 5. Association Between the Pain-Resistance Sequence and Chronicity and Between the Pain-Resistance Sequence and the Inflammatory Status of the Joint(a)

                         Pain
                        Before        Pain With       Pain After
                       Resistance     Resistance      Resistance

Chronicity versus
  pain-resistance
  sequence
Acute                      27            14              4
Subacute                   17            30              9
Chronic                    12            27              12
[chi square] = 16.10
P=0029
[V.sup.2] = .053
Inflammatory
  status versus
  pain-resistance
  sequence
Inflamed                   38            28              5
Not inflamed               18            43              20
[chi square] = 18.74
P <.00001
[V.sup.2] = .123

(a) Significance set at P < .01. [V.sup.2] = Cramer contingency coefficient.

Discussion

The results of our study provide evidence for the existence of a capsular pattern as we have defined it and for its use in identifying arthritis and arthrosis in patients with unilateral knee dysfunction. Our data do not necessarily support the existence of a capsular pattern as defined in any other way. The definition of a capsular pattern that best differentiated between subjects with and without arthritis or arthrosis appears to be a ratio of extension loss to flexion loss between 0.03 and 0.50. We believe that this definition appears to be consistent with Cyriax's original description of the capsular pattern of the knee as a "gross limitation of flexion, with slight limitation of extension."[1(p80)] Previous research examining the capsular pattern of the knee was unable to identify a proportional definition of a capsular pattern in a group of patients with osteoarthrosis of the knee? We believe that our methods, including the subject population studied, and the method of determining cutoff points defining a capsular pattern allowed for this proportional definition to emerge. We did not use Cyriax's original formulation of the pattern or those definitions suggested by other authors, but rather we developed our own formulation based on observed ratios of motion loss.

In our study, we expanded the patient population beyond those with arthrosis to include patients with more "arthritic" conditions (implying an inflammatory process) as well as patients without evidence of arthritis or arthrosis. A study attempting to examine the usefulness of a diagnostic test should include a spectrum of subjects, including those with and without the disorder that the researchers are attempting to identify.9 Without the inclusion of subjects without the target disorder, it is not possible to estimate the specificity of a test, nor is it possible to calculate likelihood ratios. It is difficult to ascertain the diagnostic usefulness of a test without these values.

As mentioned earlier, Gyriax stated that the capsular pattern of the knee joint is a "gross limitation of flexion, with slight limitation of extension."[1(p80)] He also pro vided an example: "5 or 10 degrees limitation of extension corresponds with 60 to 90 degrees limitation of flexion."[1(p600)] In a previous study examining the diagnostic utility of a capsular pattern of the knee, Hayes et al[3] adopted a definition of a capsular pattern based on the values provided in this example. Based on this definition, which corresponded to a ratio of extension loss to flexion loss between the values of 0.06 and 0.11, the authors found very few subjects with arthrosis of the knee exhibited a capsular pattern.[3] They concluded that "a proportional definition of a capsular pattern should be abandoned, but the concept of a pattern of ROM loss may be useful."[3] The authors also pointed out that had the strict proportional definition of a capsular pattern not been used, 96% of their subjects with arthrosis of the knee would have had a capsular pattern.[3]

We believe that therapists evaluating patients with knee joint dysfunction tend to interpret a capsular pattern in terms of a pattern of PROM loss, and not a strict proportional definition based on one example provided by Cyriax, but we have no evidence for this assumption. We therefore used a different approach to determining the cutoff points defining a capsular pattern in patients with knee joint dysfunction. We chose not to use the example provided by Gyriax, but instead focused on the definition of a capsular pattern as a great limitation in flexion and a slight limitation in extension. We sought to analyze the data in a manner that allowed the boundaries defining a capsular pattern to be refined prior to any statistical analysis.

We first used a histogram and a ROG curve analysis to determine the best cutoff points for defining a capsular pattern. Using this approach, a ratio of extension loss to flexion loss was identified, with sensitivity and specificity values approaching or exceeding 75% and with a positive likelihood ratio of 3.2. These values provide some evidence for the diagnostic usefulness of distinguishing between patients with and without arthritis or arthrosis. The positive likelihood ratio value found in this study means that a patient exhibiting a capsular pattern, defined as a ratio of extension loss to flexion loss between 0.03 and 0.50, is 3.2 times more likely than not to have arthritis or arthrosis of the knee. Only after these cutoff points were determined was a chi-square test of association performed. Although the results of this chi-square analysis were significant, we believe that the sensitivity, specificity, and positive likelihood ratio values are more clinically meaningful and do more to attest to the diagnostic usefulness of a capsular pattern of the knee.

The PRS is proposed to be a test of the acuity of a joint's inflammatory status and a guide to the vigor with which treatment should proceed.[1(p77)] Hayes et al[3] used chronicity (days from onset of inflammation) as a measure of acuity and found no correlation with the PRS. Other researchers[17,18] have suggested that, given the cycle of exacerbation and remission common in musculoskeletal injuries, acuity might be more accurately judged by a patient's signs and symptoms than by time from onset of inflammation. We compared both chronicity and inflammatory status with the results of the PRS. Although both chi-square statistics reached significance, the inflammatory status of the joint explained more than twice the variability in the PRS than did chronicity (12.3% versus 5.3%). Similar to Hayes et al,[3] however, we found only fair clinical agreement for the PRS (weighted [Kappa]=.28), making any conclusions regarding the validity of the PRS suspect.

We believe that this finding points to a need for further work to refine clinical judgments of the acuity of a patient's condition. One option is the development of a composite measure of several examination procedures as a basis for the judgment of acuity. Other researchers[19] have improved reliability when a composite of tests was used instead of a single test. In our study, we were able to improve the reliability of categorization of a joint as inflamed or noninflamed to a level of substantial agreement ([Kappa]=.76), whereas the individual inflammatory tests demonstrated lower levels of clinical agreement ([Kappa]= .21-.66). Another method for improving reliability is provision of additional training for clinicians and improvement of the operational definitions of terminology used in judgment of the PRS. Although we recognize the limitations of this approach with respect to generalizability, other investigators[20] have found that reliability was improved by additional training (including providing adequate operational definitions). The inflammation tests and ROM measures were not specific to this study. The inflammation tests were done as we believe Cyriax described them. What was specific to this study were the definitions of an inflamed joint versus a non-inflamed joint and of a capsular pattern versus a noncapsular pattern.

Several topics for future investigation are suggested by our study. We recognize the need to replicate the results of this study regarding the capsular pattern of the knee on other data sets and the possibility that this replication may result in further refinement of the definitions used in our study. Furthermore, the validity of measurements obtained for the other passive motion components of Cyriax's selective tissue tension scheme needs to be addressed at the knee and for other joints. The reliability and validity of measurements obtained for the PRS and of judgments of end-feel with passive motion need further examination. The best definitions of capsular patterns at other joints need to be explored.

We acknowledge several potential limitations of our study. The measurement of PROM and the calculation of ratios involve a degree of measurement error. Even though we found acceptable ICC values for these measurements (ICC=.72-.97), the standard errors of measurement demonstrate a need for cautious interpretation of the precision of these measurements. We recognize that the error inherent in the measurements used to calculate ratios of PROM loss make a strict interpretation of these ratios unwarranted. In addition, not all patients had undergone diagnostic testing; therefore, some patients may have had undiagnosed arthrosis and could have been misclassified. Furthermore, because we relied on patient-reported results of imaging studies, errors in recall or understanding on the part of the patient also may have resulted in misclassification of some patients.

We identified subjects with unilateral knee dysfunction by questioning the individual regarding prior involvement of the other knee. It is possible that some subjects may not have recalled a prior injury or may not have believed the injury to be serious enough to report. Additionally, the use of the presence of 2 out of 4 inflammatory signs as a method for categorizing a joint as inflamed has not been reported by other researchers or subjected to studies of concurrent validity with other measures of inflammatory status. We chose this technique to take advantage of the improved reliability of a composite measure as opposed to an individual test to determine inflammatory status. The use of the uninvolved limb limits the generalizability of our results to patients with unilateral knee dysfunction. We recognize that arbitrary "normal" values may need to be considered in patients with bilateral involvement.

In addition, in studies of the accuracy of diagnostic tests that use a comparison with a "gold standard," it is recommended that the individual assessing the gold standard (arthritis or arthrosis in our study) be blinded to the diagnostic test results (PROM loss).[21] In our study, the same clinician performed both tests of inflammation used in determining the gold standard and the PROM measurements. We used this method because it was not clinically feasible to have different therapists assess these 2 variables independently. The clinicians, however, were not aware that a joint would be classified as inflamed based on 2 or more positive findings, nor did they know the ratios of PROM loss that would eventually be selected as cutoff points for determining a capsular pattern.

Conclusion

This study examined 2 elements of the PROM portion of the selective tissue tension scheme described by Cyriax: the PRS and a capsular pattern of motion restriction. The PRS measurements were not found to be reliable. Examination of the capsular pattern began by examining the ratios of extension ROM loss to flexion ROM loss in those subjects who either were or were not expected to have a capsular pattern based on Cyriax's definitions. Cutoff points for the ratios of ROM loss defining a capsular pattern were determined to be 0.03 and 0.50 by examining a histogram and an ROG curve constructed from these data. Using this definition, which differs in detail but not in concept from that originally put forward by Cyriax, a capsular pattern was found to be 3.2 times more likely to be present in patients with either arthritis or arthrosis of the knee joint. Although the precision of ROM measurements taken with a goniometer make a strict interpretation of the cutoff points determined in this study unwarranted, the results provide evidence for a proportional definition of a capsular pattern of the knee as a great limitation of flexion ROM with a slight limitation of extension ROM.

References

[1] Cyriax J. Textbook of Orthopaedic Medicine, Volume 1: Diagnosis of Soft Tissue Lesions. 6th ed. London, England: Bailliere Tindall; 1982.

[2] Pellecchia GL, Paolino J, Connell J. Intertester reliability of the Cyriax evaluation in assessing patients with shoulder pain. J Orthop Sports Phys Ther. 1996;23:34-38.

[3] Hayes KW, Petersen C, Falconer J. An examination of Cyriax's passive motion tests with patients having osteoarthritis of the knee. Phys Ther. 1994;74:697-707.

[4] Reed B, Zarro VJ. Inflammation and repair and the use of thermal agents. In: Michlovitz SL, ed. Thermal Agents in Rehabilitation. 2nd ed. Philadelphia, Pa: FA Davis Co; 1990:3-17.

[5] Cohen J. A coefficient of agreement for nominal scales. Educational and Psychological Measurement. 1960;20: 37- 46.

[6] Cohen J. Weighted kappa: nominal scale agreement with provision for scaled disagreement or partial credit. Psychol Bull, 1968;70:213-220.

[7] Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull, 1979;86:420-428.

[8] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159-174.

[9] Sackett DL, Haynes RB, Guyatt CH, Tugwell P. Clinical Epidemiology: A Basic Science for Clinical Medicine. 2nd ed. Boston, Mass: Little, Brown and Co Inc; 1992:69-152.

[10] Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143: 29-36.

[11] Dujardin B, Van den Ende J, Van Compel A, et al. Likelihood ratios: a real improvement for clinical decision making? Fur J Epidemiol. 1994; 10:29-36.

[12] Simel DL, Samba GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol. 1991;44:763-770.

[13] Lacher DA. Predictive value derived from likelihood ratios: a superior technique to interpret quantitative laboratory results. Am J Clin Pathol. 1987;87:673-676.

[14] Simel DI,, Feussner JR, DeLong ER, Matchar DB. Intermediate, indeterminate, and uninterpretable diagnostic test results. Med Decis Making. 1987;7:107-114.

[15] Glass GV, Hopkins KD. Statistical Methods in Psychology and Education. 3rd ed. Boston, Mass: Allyn and Bacon; 1996:333-338.

[16] Conover WJ. Practical Nonparametric Statistics. 2nd ed. New York, NY: John Wiley & Sons Inc; 1980:178-184.

[17] Delitto A, Erhard RE, Bowling RW. A treatment-based classification approach to low back syndrome: identifying and staging patients for conservative treatment. Phys Ther. 1995;75:470-485.

[18] Von Korff M, Deyo RA, Cherkin D, Barlow W. Back pain in primary care: outcomes at 1 year. Spine. 1993;18:855-862.

[19] Cibulka MT, Delitto A, Koldehoff RM. Changes in innominate tilt after manipulation of the sacroiliac joint in patients with low back pain: an experimental study. Phys Ther. 1988;68:1359-1363.

[20] Diamond JE, Mueller MJ, Delitto A, Sinacore DR. Reliability of a diabetic foot evaluation. Phys Ther. 1989;69:797-802.

[21] Sackett DL, Richardson WS, Rosenberg W, Haynes RB. Evidence-Based Medicine: How to Practice and Teach EBM. New York, NY: Churchill Livingstone Inc; 1997:81-84.

JM Fritz, PhD, PT, ATC, is Assistant Professor, Department of Physical Therapy, School of Health and Rehabilitation Sciences, University of Pittsburgh, 6035 Forbes Tower, Pittsburgh, Pa 15260 (USA) ([e-mail]). Address all correspondence to Dr Fritz.

A Delitto, PhD, PT, is Associate Professor and Chair, Department of Physical Therapy, School of Health and Rehabilitation Sciences, University of Pittsburgh, Director of Research, Comprehensive Spine Center, University of Pittsburgh Medical Center, and Vice President for Education and Research, CORE Network, Limited Liability Corporation, McKeesport, Pa.

RE Erhard, DC, PT, is Assistant Professor, Department of Physical Therapy, School of Health and Rehabilitation Sciences, University of Pittsburgh, and Director of Physical Therapy and Chiropractic Services, Comprehensive Spine Center, University of Pittsburgh Medical Center.

M Roman, PT, is Senior Physical Therapist, Rehability Center, Raleigh, NC.

This article was submitted August 18, 1997, and was accepted May 7, 1998.