The Effect of Inter-Trial Rest Interval on the Assessment of lsokinetic Thigh Muscle Torque

Size: px

Start display at page:

Download "The Effect of Inter-Trial Rest Interval on the Assessment of lsokinetic Thigh Muscle Torque"

Moses Cook
6 years ago
Views:

1 The Effect of Inter-Trial Rest Interval on the Assessment of lsokinetic Thigh Muscle Torque PAUL W. STRATFORD, MSc,' ANNETTE BRUULSEMA,2 BILL MAXWELL,= TIM BLACK, BSc,' BEVERLY HARDING, BSc4 Journal of Orthopaedic & Sports Physical Therapy The purpose of this study was to examine the effect of two measurement protocols on the reliability of peak isokinetic knee extensor and flexor torques performed at 60 /sec. lsokinetic knee extensor and flexor torques were measured using two test protocols on 16 subjects. The two reciprocal testing protocols consisted of five trials performed with either no rest or a 30-sec rest between trials. Each subject performed both protocols with the order of protocol administration balanced across subjects. The results indicated that the Rest protocol produced average torques which were 5% greater than the No Rest protocol and that higher reliability coefficients were obtained for the Rest protocol. These findings are likely due to a significant linear trend across trials evident with the No Rest protocol. It was also demonstrated that the measurement error calculated for the average of five trials was less than that of a single trial. These findings strongly suggest that a greater measurement precision can be achieved by averaging trials obtained using a rest protocol. The assessment of knee extensor and flexor muscle strength is an activity common to clinicians and clinical researchers. Clinicians are interested in measuring strength in individual patients; clinical researchers are concerned with discerning differences between groups. Both groups are equally concerned with obtaining accurate measurements: that is, measurements which are unbiased and precise (reliable or consistent). Further, it is important that the measurements are obtained efficiently by minimizing both the testing time and the number of observations required to obtain a predetermined level of measurement precision. A review of the current literature reveals that a uniform strategy for gathering and summarizing knee extensor and flexor muscle strength data does not exist. For example, Kannus (5), examining knee extensor and flexor torques at 60 /sec and 180 /sec, selected the best of six trials (pre- 'Teaching Master. Department of Physiotherapy. Mohawk College. Assstant Professor (part-t~me). School OT & PT, McMater Ulivers~ty. Hamilton. Ontarlo. Canada. ' Physlcal therapst. Ham~lton. Ontario. Canada. "hys~cal therapy Student. Mohawk College. Physlcal therapst and cl~nical measurement consultant. Hamilton. Ont /90/ $02.00/0 THE JOURNAL OF ORTHOPAEDIC AND SPORTS PHYSICAL THERAPY Copyright by The Orthopaedic and Sports Physical Therapy Sect~ons of the American Physical Therapy Association sumably performed with no rest interval between trials) to serve as a representative sample. Kovaleski et al. (6), reporting on thigh muscle strength at velocities from 60 to 300 /sec, chose the highest peak torque value produced following at least three maximal reciprocal contractions. Epler et al. (2), examining knee extensor torques at 60 and 180 /sec, analyzed the mean torque of six successive trials. Harding et al. (3), studying peak knee extensor and flexor torques at 60 /sec designated the mean of six trials, performed with a 30 sec rest interval between each trial, to serve as the representative estimate of the subject's peak torque. An examination of these protocols reveals two major differences. The first is the practice of interposing a rest interval between trials and the second is the act of selecting a single trial-as opposed to averaging the results of a number of trials-to serve as a representation of the subject's strength. Such diversity in protocols would be unimportant to the clinician provided that these protocols all demonstrated a uniform degree of accuracy. Measurement theory would suggest that they do not (7). Kroll(7) asserts that reliability theory dictates, "If the error variances are considered to be random and uncorrelated, then the proper procedure is to use the mean of all available trials." Kroll 362 STRATFORD ET AL JOSPT 11:8 February 1990

2 further suggests that when a trend is evident, only stable trials should be averaged. A review of the current literature demonstrates few studies that have applied this theory (3, 4). Further, one may hypothesize that a declining trend in peak torque-as a result of fatigue-may exist when a suitable rest interval between trials is not incorporated into the testing protocol. Therefore, the purposes of the current study were to examine: 1) the impact of inserting a 30 sec rest interval between trials on the reliability of peak isokinetic knee extensor and flexor torques performed at 60 /sec, and 2) the impact that averaging trials had on reliability. The null hypothesis for both purposes specified no difference in reliability between the two protocols. METHOD Subjects Sixteen healthy subjects (mean and standard deviation: age 23.6 L- 2.3 years; weight, 64.6 L- 8.3 kg; 12 females) without known right knee dysfunction participated in this study. These subjects had no prior experience with isokinetic exercising. Informed consent was obtained from all subjects. Equipment The Cybexa II (Cybex Division of Lumex, Inc, Ronkonkoma, NY, ) isokinetic dynamometer and dual channel chart recorder were used to obtain the data. Testing was performed at 60 / sec. The chart recorder paper speed was 5 mm/ sec and a damp setting of 2 was used. The peak torques were abstracted by hand using the torque template provided by Cybex. Procedure All subjects performed two measurement protocols: 1) no rest between trials (NR), and 2) a 30- sec rest interval between trials (R). Randomization was performed in pairs to ensure that an equal number of subjects were assigned to each protocol order presentation (i.e., eight subjects were tested in each of the following orders: NR-R and R-NR). Subjects were positioned on the testing bench and stabilization straps were secured over the chest, pelvis, and distal aspect of the right thigh. The lateral epicondyle of the knee was visually aligned with the axis of rotation of the dynamometer. The distal aspect of the input adaptor pad was positioned 2 cm proximal to the tibiotalar joint line. Knee range of motion was standardized for each patient. The patients started with their knees flexed such that their heels were touching the flexion limiting pad. This was followed by complete extension to anatomical zero and then complete flexion until the heel once again touched the pad. The warm-up consisted of easy pedaling on a stationary bicycle for 5 minutes, followed by a 5-minute rest interval. The subjects then performed four progressive warm-up trials at 60 / sec (the first at half-effort, followed by two at three-quarter effort, and one at full effort). A 2- minute rest interval was then observed prior to testing at this point and during the interval between protocols. The No Rest protocol consisted of having the subjects perform five maximum reciprocal extension-flexion trials. The Rest protocol consisted of having the subjects perform five maximum reciprocal trials, each being separated by a 30-sec rest interval in flexion. Both protocols included a 1 -sec hesitation in full extension that was actively maintained by the subject. Subjects were verbally prompted to initiate the movement, but were not encouraged during the contraction period. No patient feedback was provided either among trials or between protocols. This was identical for both protocols. Analysis Means and standard deviations were calculated for each protocol, trial, and order (i.e., first test versus second test regardless of protocol). Differences in peak torque due to protocol and order were formally tested by means of a three-way analysis of variance (the factors were subjects, order, and protocol); whereas differences due to patients and trials were evaluated using a twoway analysis of variance. When statistically significant differences were identified for the Trial factor, a trend analysis was also performed. Finally, the equality of the overall within subject error variances (trial and error combined: this being the equivalent of a one-way analysis of variance error term) for each muscle group was tested using an F test. Also, intraclass correlation coefficients were calculated for a single measurement (R,) and the mean of five measurements (R,): these coefficients represent Shrout and Fleiss (9) classifications (2,l) and (2,5). Further, 95% confidence limits were calculated for R, (9). In order that the measurement error could be expressed in clinically meaningful terms (i.e., expressed in newton meters), the standard error of measurement (SEM) was also calculated (10). (The SEMs presented in this paper are derived from a one-way ANOVA configuration. Here, the SEM is equal to the JMTIJN, where MSE represents the mean square error term and N represents the number of trials.) For all tests, statistical significance was declared if the p value was less than JOSPT 1 1:8 February 1990 EFFECT OF INTER-TRIAL REST INTERVAL 363

RESULTS TABLE 4 Extensor Rest protocol analysis of variance table Table 1 provides the mean scores for the categories Protocol and Order.

The Trial mean scores are presented in Table 2 and the corresponding analysis of variance tables are provided in Tables 3-6.

When the Rest protocol was used, no significant difference among trials was identified for either muscle group (i.e., no trend was present). TABLE 1 Summary of peak torques (N.

t Standard deviation. TABLE 2 Summary of peak torques (N.m) by trial Extension No Rest 167.5' 39.37 Rest 162.4 45.6 Flexion No Rest 95.0 25.9 Rest 94.9 26.1 ' Mean. t Standard deviation.

6' (Linear trend) 1 4841 4841 78.3' Error 60 3710 62 ' p < 0.001. Swrce dl SS MS -- -..-... F Among subjects 15 147200 9813 187.7' Withinsubjects 64 3187 50 Trials 4 48 12 0.

6' (Linear trend) 1 1074 1074 30.3' Error 60 2124 35 ' p < 0.001. TABLE 6 Flexor Rest protocol analysis of variance table Source df SS MS F Among subjects 15 51633 3442 172.

3 RESULTS TABLE 4 Extensor Rest protocol analysis of variance table Table 1 provides the mean scores for the categories Protocol and Order. The three-way analysis of variance revealed that statistically significant differences (p < 0.001) exist for the Protocol category, but not for Order. The Trial mean scores are presented in Table 2 and the corresponding analysis of variance tables are provided in Tables 3-6. By referring to Tables 3 and 5 it is evident that a significant linear trend exists for the knee extensor and flexor muscle groups for the No Rest protocol. When the Rest protocol was used, no significant difference among trials was identified for either muscle group (i.e., no trend was present). TABLE 1 Summary of peak torques (N. m) by Protocol and Order Muscle Group Extensor Flexor.- Protocol No Rest 156.4' t 24.6 Rest Order First Test Second Test Mean. t Standard deviation. TABLE 2 Summary of peak torques (N.m) by trial Extension No Rest 167.5' Rest Flexion No Rest Rest ' Mean. t Standard deviation. Trial TABLE 3 Extensor No Rest protocol analysis of variance table Source df SS MS F Among subjects ' Within subjects Trials ' (Linear trend) ' Error ' p < Swrce dl SS MS F Among subjects ' Withinsubjects Trials Error TABLE 5 Flexor No Rest protocol analysis of variance table Source dl SS MS.- p- -. F Among subjects ' Within subjects Trials ' (Linear trend) ' Error ' p < TABLE 6 Flexor Rest protocol analysis of variance table Source df SS MS F Among subjects ' Within subjects Trials Error TABLE 7 Error estimates for one and five trials for the No Rest and Rest protocols Extensor values R SEM 95% C.I. for SEMt Flexor values R SEM 95% C.I. for SEMt trial No Rest Mean of 5 tr~als trial Rest Mean of 5 tnals - - The combined within subject error variances (trials plus error) for both the extensor and flexor data indicate that the No Rest variance was 2.7 and 2.5 times greater than the Rest variance (p < 0.001), respectively. An examination of the reliability coefficients presented in Table 7 indicate two things. First, the coefficients for the Rest protocol are greater than those for the No Rest protocol (the 95% confidence limits on R, for the Rest and No Rest protocols are and , respectively); second, the coefficients for the mean of five trials are greater than those based on a single 364 STRATFORD ET AL JOSPT 11:8 February 1990

4 trial. This finding is also re-enforced by the SEM. These results, expressed in newton meters, indicate that the error associated with a single measurement was greater than that for the average of five measurements and that the error associated with the No Rest protocol was greater than error specific to the Rest protocol. DISCUSSION Several findings relevant to strength measurement in general, and as they pertain to the practice of physical therapy in particular, will be discussed. Also, suggestions for future research will be offered. The first finding comments on measurement accuracy. Recalling the introduction, an accurate measurement was defined as one which is unbiased and precise. The current work demonstrates that the mean peak torque-both flexor and extensor-for the No Rest Protocol tends to be less than those of the Rest protocol by about 5%. This finding is due to the declining linear trend in peak torque which is associated with the No Rest protocol. This phenomenon has been identified previously by Barnes (1) (who examined isokinetic knee extensor torque) and by Patton et al. (8) (who examined isokinetic elbow flexor torque), who described it as a fatigue curve. By referring to Table 2, it is evident that the knee extensor torque for the fifth trial was 87% of that of the first trial. The corresponding value for the hamstrings was 90%. The extensor finding is in agreement with the work of Barnes (I), who also demonstrated a value of 87% when testing the knee extensors at 60 /sec (Barnes did not study flexor torques). In addition to producing a bias, the No Rest protocol also has a significant impact of measurement precision. This was verified by demonstrating that the error variance associated with the No Rest protocol was significantly greater than that of the Rest protocol. While a fatigue effect may be obvious to clinicians who frequently observe this phenomenon, it has significant impact on measurement error, an impact which has been generally overlooked. Guidelines for reducing this error have been introduced by Kroll (7). He suggests that measurement theory dictates that when a trend is evident, only stable trials should be averaged. But few studies have applied this theory (3, 4), probably because the impact of this error has been underestimated by clinicians. In order to demonstrate the clinical importance of Kroll's advice, the following example is offered. Typically, the clinician assesses the patient, identifies whether the patient requires treatment, provides treatment when appropriate, reassesses the patient at some subsequent point, and dis- charges the patient when the goals of treatment have been met. Inherent in this statement are two key points, the ability to differentiate those patients requiring treatment from those who do not and the ability to detect a change in a patient over time. The intraclass correlation coefficient (R), presented in Table 7, is an index of the test protocol's ability to differentiate among subjects. Thus, within the context of strength measurements, an increase in R results in an increased ability to differentiate those patients requiring strength training from those who do not. Intrinsic to the concept of detecting change in a patient's strength over time is the notion that the observed strength change must be greater than the magnitude of error associated with the measurement. The SEM (more specifically, the corresponding 95% confidence limits) provides the theoretical lower boundary of detectable change. For example, if at the initial assessment a clinician determined a patient's knee extensor strength to be 100 N. m based on a single trial that was selected from a series of trials sampled using the No Rest protocol, one would be 95% certain that the patient's true strength at that occasion would fall between 77.3 and N m (i.e., 100 a 22.7 N. m, from Table 7). Accordingly, if the patient's strength was based on the mean of five trials obtained using the Rest protocol, one would be 95% certain that the patient's true strength would lie between 93.7 and N-m (i.e., 100 k 6.3 N. m). In the first example the patient would have to demonstrate a strength gain in excess of 22.7 N. m before the clinician could be 95% certain that a true change occurred; in the second example a change exceeding 6.3 N.m would be required. Thus, the practice of eliminating a trend and averaging a number of trials improves the clinician's ability to differentiate among subjects and in detecting changes within patients over time. The findings expressed above provide direction for future research involving three variables, testing velocity, duration of rest interval, and the number of trials required to meet a predetermined level of reliability. For example, the testing velocity for this study was 60 /sec and all five trials performed using the Rest protocol were found to be stable. Johnson and Siegel (4), using a Rest protocol and testing at 1 80 /sec, noticed an increasing trend in peak torque over the first three of six trials. A causal hypothesis for this difference may be related to the difference in testing velocity. With respect to the duration of rest interval, further work will be required to determine the optimal (minimal) rest interval required to obliterate a trend. It is conceivable that the appropriate rest interval time may be a function of the testing velocity. Finally, it is also conceivable that the total number of trials that the patient needs to perform may vary. For example, the current work was able JOSPT 1 1:8 February 1990 EFFECT OF INTER-TRIAL REST INTERVAL 365

5 to make use of all five trials that were performed, while Johnson and Siegel (4) could only average the last three trials. Thus, if the average of five trials were required to achieve the desired level of reliability, only five trials would be required at 60 / sec compared to eight at 1 80 /sec (the first three being eliminated due to the trend). CONCLUSION This study clearly reveals three findings: 1) that the mean flexor and extensor torques estimated from the No Rest protocol are 5% less than the Rest protocol estimate of the same torques, 2) the measurement error associated with the No Rest protocol is greater than that of the Rest protocol, and 3) that the reliability based on the average of a number of values is greater than that of a single measurement. The current work, when considered along with the existing literature, suggests that subsequent research be undertaken to examine the interaction of testing velocity with the duration of rest interval and the number of trials required to obtain a reliability coefficient of acceptable magnitude. 0 REFERENCES Barnes WS: lsokinetic fatigue curves at different contractile velocities. Arch Phys Med Rehabil 62: Epler M. Nawoczenski D. Englehardt T: Comparison of the Cybex II standard shin adapter versus the Johnson anti-shear device in torque generation. J Orthop Sports Phys Ther 9: Hardmg B. Black T. Bruulsema A. Maxwell 8. Stratford P: Reliab~lity of a reciprocal test protocol performed on the Kmetic Communlcator: an isokinettc test of knee extensor and flexor strength. J Orthop Sports Phys Ther 10: Johnson J. S~egel D: Reliability of an isokmetic movement of the knee extensors. Res O 49: Kannus P: Peak torque and total work relationship in the thigh muscles after anterlor cruclate lhgament injury. J Orthop Sports Phys Ther 10: Kovalesk~ JE. Craig BW. Costill DL. Habansky AJ. Matchen W: Influence of age on muscle strength and knee function following arthroscopic meniscectomy. J Orthop Sports Phys Ther 10: Krdl W: Reliability theory and research decision in selection of a crtterion score. Res 0 38: Patton RW. Hinson MM. Arnold Jr BR. Lessard 6: Fatigue curves of isokinetic contractions. Arch Phys Med 59: Shrout PE. Fleiss J: lntraclass conelattons: uses in assessing rater reliab~lity. Psychd Bull 86: Verducci FM: Measurement Concepts In Physical Education, pp Toronto. Ont: CV Mosby STRATFORD ET AL JOSPT 11:8 February 1990

The Reliability of Four Different Methods. of Calculating Quadriceps Peak Torque Angle- Specific Torques at 30, 60, and 75

The Reliability of Four Different Methods. of Calculating Quadriceps Peak Torque Angle- Specific Torques at 30, 60, and 75 By: Brent L. Arnold and David H. Perrin * Arnold, B.A., & Perrin, D.H. (1993).