Reliability of measurements obtained with four tests for patellofemoral alignment.

The term "patellofemoral (PF) malalignment" implies that there is a disruption in the normal tracking of the patella in the femoral groove during knee motion. [1] This condition may result in abnormal stresses being applied to structures associated with the PF joint, producing pain and inflammation. A number of physical therapy interventions have been described for PF malalignment, all of which emphasize reducing factors that may contribute to the malalignment. [1-7]

McConnell [4] has described a series of tests for determining PF alignment. These tests examine the presence or absence of medial/lateral displacement, medial/lateral tilt, medial/lateral rotation, and anterior tilting of the patella with respect to the femur. The results of the tests are used to assist the therapist in making treatment decisions regarding PF taping techniques and therapeutic exercise procedures, which are also described by McConnell. [4]

The reliability of measurements obtained with the PF alignment tests described by McConnell [4] has not been reported. A preliminary investigation performed by the first author (GKF) on 30 subjects, with and without PF pain, suggested that intratester reliability of these measurements was poor. If the reliability of measurements obtained with this evaluation procedure are poor, clinical decisions that are made based on this procedure may not be valid.

A number of consequences may result from selecting treatment for PF dysfunction function from unreliable measurements of PF alignment. For example, if PF taping techniques are selected based on unreliable measurements, the treatment may not be effective. Taping may then be incorrectly dismissed as an ineffective treatment for a patient who may be helped by a different taping technique. Unreliable measurements of alignment may also lead to a false impression about the mechanism involved if taping is successful in relieving symptoms. For example, suppose we have concluded that our patient has an excessive literal displacement of the patella based on our examination of alignment. If the measurement is unreliable, there is a chance that perhaps the patella is not displaced or displaced medially. Nevertheless, we would select a taping technique for a lateral displacement. If our patient's condition improves, we may assume it was because we restored normal PF joint alignment when in fact the improvement had nothing to do with joint alignment.

Based on the results of the preliminary investigation, a broader study including more subjects and testers at several clinical facilities was deemed necessary to examine reliability. The purpose of this study was to determine the intertester reliability of measurements taken with four tests for PF alignment, as described by McConnell. [4] We chose to design the study in a similar fashion to other investigators assessing reliability of clinical assessments. [8]

Method

Subjects

Sixty-six subjects (31 male, 35 female) participated in the study. Subject ages ranged from 14 to 74 years (X [bar][+ or -]SD=29.7 [+ or -]13.1). Mean height and weight were 171.2 [+ or -]10.2 cm (67.4+/-4.0 in) and 73.4+/-19.6 kg (161.9 [+ or -]43.1 lb), respectively. All subjects were referred for physical therapy by their physicians. Subjects were included in the study if the examining therapist believed that assessment of PF alignment would be part of the typical physical therapy evaluation for that patient. Forty subjects had diagnoses directly related to PF dysfunction (ie, PF pain syndrome, anterior knee pain, chondromalacia patellae, subluxating patella, patellar tendinitis, patellar fracture). The remaining subjects had diagnoses consisting of meniscal pathology, ligamentous pathology, and fractures of the femur or tibia. Subjects were excluded from the study if they had received a surgical procedure specifically to realign the patella (eg, lateral retinacular release). All subjects signed an informed consent form prior to participation in the study.

Testers

Testers in this study were 12 physical therapists employed at one of four physical therapy clinics in the Philadelphia, Pa, area. All testers frequently treated patients with knee disorders or PF joint dysfunction in their practice. All testers used PF taping techniques for treating these patients and were familiar with the PF alignment tests prior to participation in the study. Only 1 tester learned the alignment tests from attending a continuing education course given by McConnell. All other testers learned the alignment tests by reading published descriptions of the tests [4] or from colleagues who learned the tests in other continuing education courses. Table 1 provides demographic information of therapists at each clinic site.

Table 1. Characteristics of Testers
         Knee                          Therapist No.
         Disorders (a)   Patients      (Years of
Clinic   Case Load       Examine (b)   Experience) (c)
A        60              19            1 (4.5)
                                       2 (1.5)
B        20              13            1 (7.0)
C        25              22            1 (17.0)
D        20              12            1 (4.5)
(a) Percentage of the therapists' case loads that comprised
patients with
knee disorders. (b) The number of patients examined for this study
from each
clinic. (c) The experience (in years) of each therapist treating
patients
with knee disorders.

Patellofemoral Alignment Tests

Because the testers participating in the study learned the PF alignment tests from varying sources, we decided to provide standardized instructions of these tests. All testers received a written description and a photograph of each specific test of alignment. We chose this method of standardizing the instructions because it would allow our results to be generalized to therapists who followed our written instructions and photographs of the tests. We elected not to perform personal instruction or demonstrations of the tests, because then the results would only be generalizable to therapists who received our personal instruction. The following instructions for the alignment tests are based on descriptions from McConnell. [4] The wording in following descriptions is exactly as they were presented to the testers.

Medial/lateral displacement. Lateral displacement is determined by palpating the medial and lateral femoral epicondyles with the index fingers and simultaneously palpating the midpatella with the thumbs (Fig. 1). Normally, the distance between the index fingers and the thumbs should be approximately the same. [4] If a lateral displacement is present, then the distance from the index finger palpating the lateral epicondyle to the thumbs will be less than the distance from the finger palpating the medial epicondyle to the thumbs. [4] If a medial displacement is present, the distance from the medial epicondyle to the thumbs will be less than the distance from the lateral epicondyle to the thumbs. [4]

Medial/lateral tilt. The degree of medial or lateral patellar tilting is determined by comparing the height of the medial patellar border with that of the lateral patellar border. The examiner places his or her thumb and index finger on the medial and lateral borders of the patella (Fig. 2). Both digits should be of equal height. If the digit palpating the medial border is more anterior than the lateral border, then the patella is tilted laterally.4 If the digit palpating the lateral border is more anterior than the medial border, then the patella is tilted medially. [4]

Anterior tilt. Anterior tilt is determined by palpating the inferior pole of the patella (Fig. 3). If no significant anterior tilt exists, the inferior pole should be easily palpated. [4] An anterior tilt is present if the examiner must place a downward pressure on the superior pole of the patella so that the inferior pole becomes superficial enough to palpate. [4]

Patellar rotation. Patellar rotation is determined by examining the relationship between the longitudinal axis of the patella and the longitudinal axis of the femur (Fig. 4). The longitudinal axis of the patella should normally be in line with the anterior superior iliac spine (ASIS). If the distal end of the longitudinal axis of the patella is angled lateral to the ASIS, then the patella is considered to be rotated laterally. [4] If the distal end of the longitudinal axis of the patella is angled medial to the ASIS, then the patella is considered to be rotated medially. [4]

Experimental Procedure

All therapists serving as testers received the written instructions and photographs of the PF alignment tests approximately 2 weeks prior to data collection. This procedure allowed testers the opportunity to practice the evaluation procedures prior to participation in the study. We provided testers with the opportunity to ask questions regarding interpretations of the written instructions on the day that testing was initiated at each clinic.

Initially, an attempt was made to select a random pair of therapists to test each patient. Random pairing was not maintained, however, due to scheduling conflicts for both testers and patients. One facility had only two therapists participating in the study, so random pairing was not possible.

Paired testers performed the PF alignment tests independently. The therapist assigned to treat the subject at the time of referral to physical therapy was identified as examiner 1. The paired therapist was identified as examiner 2. Examiner 1 performed the tests of PF alignment first, and then examiner 2 performed the tests within one treatment session of examiner 1. Most often, testing was performed by both testers during the same session.

The testers were instructed not to discuss the evaluation findings on any subjects until the entire study was completed. This instruction was given to prevent a tester from obtaining test results or information about the examination process from the other testers.

One individual from each participating facility was designated as the data-collection coordinator. This person collected examination result forms from the examiners, placed the forms in an envelope, and returned these forms to us. This procedure was done in an attempt to maintain confidentiality of the examination results.

Data Analysis

Intertester reliability was determined by calculating kappa coefficients, which are appropriate for nominal-level data. [9] The kappa coefficient is based on the percentage of agreement between repeated assessments that has been corrected for chance agreement.

Results

There were a total of 66 paired assessments for each test of PF alignment. The percentages of agreement and kappa coefficients are reported in Table 2. The percentages of agreement ranged from 440/o to 71%. The kappa coefficients ranged from .10 to .36.

Table 2. Kappa Statistic (Percentage
of Agreement in Parentheses) for Patellofemoral
Alignment Evaluation
Procedures
                             Kappa
                             (Percentage of
Procedure                    Agreement)
Medial/lateral
  displacement               .10 (44%)
Medial/lateral tilt           .21 (59%)
Anterior/posterior tilt      .24 (71%)
Medial/lateral
  rotation                   .36 (61%)

Discussion

The low kappa coefficients suggest the reliability of measurements of PF alignment ranged from poor (.00-.10) to fair (.20-.40), according to criteria proposed by Landis and Koch. [10] The kappa coefficients were consistently lower than the percentage-of-agreement values because kappa coefficients represent the proportion of agreement after chance agreement is removed.

The kappa coefficient may be artificially lowered or elevated if there is insufficient variability in the phenomena being assessed. [11] This artificially lowered or elevated value may result in either an underestimation or overestimation of reliability. For example, in the evaluation of medial/lateral displacement, there are three possible choices: medial displacement, lateral displacement, or no displacement. If the sample consisted of subjects who were equally distributed among each of the three possible test results, then the sample would demonstrate reasonable variability in the phenomena being studied and the kappa coefficient would be appropriate. However, if almost all subjects in the sample exhibited a lateral displacement, then the sample would lack adequate variability. A lack of variability would result in a high degree of agreement being attributed to chance and, therefore, a low kappa coefficient.

According to Feinstein and Cicchetti, [11] there is no "gold standard" by which adequate variability of scores is determined when using kappa. They suggest that analysis of the marginal totals of scores can indicate whether kappa will be artificially elevated or lowered. If the marginal totals are imbalanced in their distribution, then kappa will be altered. If marginal totals are symmetrically imbalanced, then kappa will be lowered and win tend to underestimate reliability. If marginal totals are asymmetrically imbalanced, kappa will be elevated and will tend to overestimate reliability. The effect of marginal total imbalances on the kappa statistic is probably most serious in cases in which a high percentage of agreement exists.

The distributions of judgments for each test of PF alignment are shown in Tables 3 through 6. The marginal totals (in italics) indicate that there is a symmetrical imbalance (vertical totals increase from top to bottom, horizontal totals increase from left to right) for each table [11], thus, our kappa values may be artificially lowered. We do not, however, consider the percentage of agreement to be high for any of the tests, and therefore kappa would probably not be seriously affected. Hence, we believe our kappa values reflect the degree of reliability of the PF alignment measurements, and they are poor to fair at best.

Table 3. Distribution of Paired Judgments for
Medial/lateral Displacement
        Displacement
        Medial   None    Lateral   Total
Medial    5        3       3        11
None      4       10      10        24
Lateral   3        9      19        31
Total     12      22      32
Table 4. Distribution of Paired Judgments
for Medial/lateral Tilting
        Tilting
          Medial   None   Lateral   Total
Medial    0         0      1          1
None      3        11     11         25
Lateral   3         9     28         40
Total     6        20     40
Table 5. Distribution of Paired Judgments
for Medial/lateral Rotation
       Rotation
         Medial   None   Lateral   Total
Medial   2          1       6        9
None     3         17       5       25
Lateral  4          7      21       32
Total    9         25      32
Table 6. Distribution of Paired Judgments
for Anterior Tilting
           Tilting
           Anterior   None   Total
Anterior      7         8     15
None         11        40     51
Total        18        48

Several factors may have contributed to the poor reliability of the PF alignment measurements in this study. Perhaps the most influential factor is that tester's were required to palpate bony landmarks that may be difficult to palpate accurately. The examination procedure for medial/lateral displacement serves as an example. This procedure involves palpation of the medial and lateral femoral epicondyles and the center of the patella. The medial and lateral femoral epicondyles are not discrete prominences, and they vary in size and shape across individuals. It is possible that two examiners could correctly palpate one of these structures but their finger placements may be several millimeters apart. Likewise, the shape, size, and orientation of the patella varies across individuals, making it difficult to accurately palpate the center of the patella.

Errors in visual inspection of the patellar position with respect to the femur may have contributed to the poor reliability of the measurements. Because the bony segments of the PF joint are relatively small, errors due to visual estimation may have been magnified.

All examiners were given the same written instructions and photographs of the evaluation procedures. We believed that this was an acceptable way of standardizing the testing procedures, because many therapists learn evaluation and treatment techniques through descriptions of the techniques in the literature. It may be possible that these instructions were interpreted differently across therapists. We attempted to minimize problems with interpretation of the instructions by providing a question-and-answer period on the day that testing was initiated at each clinic. This precaution would not, however, completely eliminate the potential for varying interpretations of the instructions. The written instructions given to the examiners were based on the descriptions of the procedures provided by McConnell. [4] Modifications to these instructions may improve the reliability of the evaluation procedure. We believe, however, that an evaluation that is based on palpation of nondiscrete bony landmarks and visual inspection of relatively small bony segments is inherently unreliable.

Another potential source of error was that examiner 1 always had knowledge of the physician's diagnosis for the patient because examiner I was the treating therapist. Unfortunately, we are not certain whether examiner 2 always had this information. Therefore, examiner 1 may have been biased by knowing the physician's diagnosis prior to performing the PF alignment tests.

Clinical Implications

The PF alignment tests in this study were initially developed, in part, to guide clinical decisions regarding PF taping procedures. Because these PF alignment tests do not exhibit a high degree of reliability, they may not provide valid information that would guide clinical decisions for PF taping or therapeutic exercise. This concern does not, however, discount the usefulness of PF taping techniques. in our clinical experience, the PF taping procedures described by McConnell [4] can be effective in reducing symptoms.

Physical therapists may choose to use other methods of determining when and how to apply PF taping techniques. Some clinicians in our area use the patellar tilt test and patellar glide test described by Kolowich et al [12] to guide clinical decisions regarding taping. The reliability of measurements obtained with these tests, however, is also unknown.

Because we have been unable to find a reliable clinical method of assessing PF alignment, we recommend selecting taping techniques based on the immediate response of the patient's symptoms to a specific taping technique. This method is not dependent on any judgment of PF alignment. The critical first step in this method is to identify a painful activity, such as stair climbing, squatting, or manually resisted knee extension. This activity serves as the basis for judging the effect of the tape on the patient's symptoms. We then systematically apply specific taping techniques until the patient's symptoms are significantly reduced or completely alleviated.

Conclusions

Four clinical tests of PF alignment were found to lack reliability in this study. Without reliability, these tests would not be useful in guiding treatment decisions for PF taping or therapeutic exercise. Therapists are encouraged to seek other methods to guide clinical decision making for these treatments.

Acknowledgments

We thank the physical therapy staffs at the following clinics in the Philadelphia metropolitan area: Temple University Sportsmedicine Clinic, South jersey Physical Therapy Associates, Pennsylvania Rehab Inc, and Atlantic Rehabilitation Services. We extend special thanks to Jeff Ryan, PT, ATC, Ned Lenny, PT, Scott Voshell, PT, and Deborah Tullman, PT, for their efforts in coordinating data collection at the participating facilities.

References

[1] Kramer PG. Patella malalignment syndrome: a rationale to reduce excessive lateral pressure. J Orthop Sports Phys Ther. 1986;8:301-309. [2] Fisher RL. Conservative treatment of patellofemoral pain. Orthop Clin North Am. 1986; 17:269-272. [3] Henry JH. Conservative treatment of patellofemoral subluxation. Clin Sports Med. 1989; 8:261-278. [4] McConnell J. The management of chondromalacia patellae: a long-term solution. Australian Journal of Physiotherapy. 1986;32:215-223. [5] Paulos L, Rusche K, Johnson C, Noyes FR. Patellar malalignment: a treatment rationale. Phys Ther. 1980;60:1624--1632. [6] Shelton GL, Thigpen LK. Rehabilitation of patellofemoral dysfunction: a review of literature. J Orthop Sports Phys Ther. 1991;14:243-249. [7] Woodall W, Welsh J. A biomechanical basis for rehabilitation programs involving the patellofemoral joint. J Orthop Sports Phys Ther. 1990;11:535-541. [8] Riddle DL, Rothstein JM. Intertester reliability of McKenzie's classifications of the syndrome types present in patients with low back pain. Spine. 1993;18:1333-1344. [9] Cohen J. A coefficient of agreement for nominal scales. Educational and Psychological Measurement. 1960; 20:37- 46. [10] Landis RJ, Koch GG. The measurement of observer agreement for categorical data. Biometrics. metrics. 1977;33:159-174. [11] Feinstein AR, Cicchetti DV. High agreement but low kappa, 1: the problems of two paradoxes. J Clin Epidemiol. 1990;43:543-549. [12] Kolowich PA, Paulos LE, Rosenberg TD, Farnsworth S. Lateral release of the patella: indications and contraindications. Am j Sports Med. 1990;18:359-365.

GK Fitzgerald, PT, OCS, is Assistant Professor, Department of Physical Therapy, Hahnemann University, Mailstop 502, Broad and Vine Sts, Philadelphia, PA 19102 (USA). Address all correspondence to Mr. Fitzgerald.

PW McClure, PT, OCS, is Assistant Professor, Department of Physical Therapy, Hahnemann University.