Journal of Orthopaedic & Sports Physical Therapy 2OOO;3O (4):170-1 82
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
Influence of Ankle Support on Joint Range of Motion Before and After Exercise: A Meta-Analysis Mitchell 1. Cordova, PhD, ATC1 Christopher D. Ingersoll, PhD, ATC, FACSM * Michael 1. LeBlanc, PhD
t has been estimated that 1 million people encounter acute ankle injuries every ~ear.~%kle sprains have been identified as the most common injury occurring during athletic parti~ipation.~~~~" Because the incidence of ankle injury is high, epidemiological research has been conducted to characterize the role of various ankle support devices in reducing ankle injuryw.JsJ%nd injury frequency r a t e ~ . 2 . ~ Ankle . ~ - ~bracing . ~ ~ and taping are effective in reducing ankle injuries in sports because of the increased structural support provided by these device^.^.^^+^^.^" The main function of these appliances is to restrict frontal plane movement of the subtalar joint. However, application of adhesive taping and lace-up braces are also designed to limit ankle plantar flexion. The effects of ankle taping, lace-up braces, and semirigid orhoses on inversion12-lR,2V2F~2UeJI ,.% and ever~ion~~-'*:~":~~ range of moKey Words: ankle bracing, ankle injury, joint motion, meta-analysis tion (ROM) before and after exerAssistant professor and director, Sports Injury Research laboratory, Athletic Training Department, cise have been extensively studied. Indiana State University, Terre Haute, Ind. To a lesser degree, the effect of Professor and chairperson, Athletic Training Department, Indiana State University,Terre Haute, Ind. such devices on plantar flexion Assistant professor, Department of Counseling and Psychological Services, S.U.N.YGwego, Os- ROM (PROM)24.2527.51.%and dorsiwego, NY beSend correspondence to Mitchell 1. Cordova, Sports Injury Research Laboratory, Athletic Training flexion ROM (DROM)z4.2627 fore and after exercise has also Department, lndiana State University, Terre Haute, IN 47809. E-mail:
[email protected]
Study Design: Meta-analysis. Objective: To evaluate the effects of different types of ankle support on ankle and foot joint range of motion before and after activity using meta-analysis procedures. Background: The effects of ankle support on joint range of motion before and after exercise has been extensively studied, but the results among studies are not consistent. Obtaining knowledge from synthesizing the available literature with a meta-analysis can provide a greater understandingof these effects. Methods and Measures: A total of 253 cases from 19 studies were examined and included in this analysis. The treatment variables were ankle support with 3 levels (tape, lace-up, and semirigid) and time with 2 levels (before exercise and after exercise). Standardized effect sizes were computed for inversion, eversion, dorsiflexion, and plantar flexion range of motion to measure the difference between control and treatment groups at each point in time. Effect sizes were analyzed using a mixed-model factorial analysis of variance. Resulk Before exercise, the semirigid condition (-2.97 0.63) demonstrated greater restriction compared with the tape (-2.33 0.38) and lace-up conditions (-2.18 2 0.86) for inversion 0.64) restricted inversion range range of motion. After exercise, the semirigid condition (-3.85 of motion more than the tape (-1 .O7 0.20) and lace-up (-1.56 0.29) conditions. No differences were found between the mean elkt sizes for the tape and lace-up conditions before and after exercise. With respect to eversion range of motion, the semirigid support (-2.69 + 0.43) provided greater restraint compared with the tape (-1.00 + 0.21) and lace-up (-1.40 + 0.47) conditions. The laceup condition also displayed greater support compared with tape alone. For dorsiflexion range of motion, greatest overall support was provided by the tape condition (-0.94 0.06) compared with the laceup condition (-0.51 + 0.06). Conclusions: The greatest restriction of motion in the frontal plane was offered by the semirigid support condition, whereas taping offered the most support for limiting dorsiflexion range of motion. The results of this study may help clinicians make rational decisions concerning the selection of ankle appliances for preventing acute or chronic reinjury. / Orthop Sports Phys Ther 2OOO;3O: 170-182.
+
+
+
+
+
+
I
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
been evaluated. However, the results of these comparative studies are not all in agreement. This is partly due to varied techniques used in measuring ROM, different applications of adhesive tape, and different styles of commercially available lace-up braces and semirigid orthoses. Other sources of inconsistencies in the literature result from studies with inadequate research designs and lack of statistical power. The general effectiveness of ankle bracing and taping has been discussed in 2 recent review^.^"^^ Unfortunately, these reviews relied on a narrative approach, where outcomes of various studies were merely described and not statistically analyzed. Meta-analysis is a statistical procedure that combines the results of related studies in a systematic and comprehensive manner. By synthesizing results across studies, meta-analyses attempt to answer ques tions related to the generalizability of study results to the population. Unlike traditional experimental or quasiexperimental research, meta-analyses use statis tical results from individual studies as the units of measurement in an overall study.1° Although individual research studies provide insight into true relationships, aggregation of accumulated findings more effectively increases knowledge by providing greater synthesis of information. Furthermore, it allows for research findings to be generalized to groups beyond the original samples. By considering replicated studies in one overall study, meta-analyses address the alltoocommon phrase found in many research articles: "more studies are needed." To our knowledge, no comprehensive statistical analysis of the literature concerning the effectiveness of ankle support on ankle and foot ROM has been reported. Thus, the purpose of this study was to use a meta-analytic a p proach to evaluate ankle support effects on ankle and foot ROM before and after exercise.
ing which studies would be included in the metaanalysis, the nature of the experimental hypotheses was not considered. It was felt that, regardless of what results the authors would find in their individual studies, this information is irrelevant to statistically quantifying the effects from these studies. The investigators were not blind to the findings in each study when decisions were made to reject or accept the article for analysis. To avoid bias in determining which studies would be included in our analysis, each individual study's information was evaluated from our code sheet. If any of the studies failed to meet all our inclusion criteria, they were not incorporated in the analysis. Primarily, studies were not included in our analysis because SDs were not reported with their means. Therefore, effect sizes could not be computed. Studies that used subjects with chronically injured ankles were excluded from this analysis. It needs to be emphasized that the ankle brace and time effects from the individual studies were extracted to create the effects sizes in this study. In other words, not every study evaluated each type of ankle support nor did every study evaluate the effect of time on ROM. With respect to the reliability of the data used in the present analysis, approximately 11 (48%) of the 19 studies reported reliability coefficients concerning their measurements. Because of this low percentage, we did not include reporting reliability estimates as a criteria for study inclusion. However, of the studies that reported reliability coefficients used in this analysis, the test-retest reliability varied from 0.92-0.99 for inversion, 0.75-0.99 for eversion, 0.70-0.95 for plantar flexion, and 0.92-0.95 for dorsiflexion. Because many of the studies were similar in design and measurement device used, it was assumed that the other studies would have similar reliability estimates concerning these variables.
METHODS
Coding of Variables
All studies used in the analysis had the following descriptive information coded: age, height, mass, method of ROM measurement, total number of s u b The literature reviewed for this study was limited jects used in the study, and number of subjects per to published studies reported between 1966 and experimental condition. The preliminary dependent 1997. Studies were obtained from MEDLINE and measures used were inversion ROM (IROM), everSports Discus database searches using the following key words: ankb bracing, ankle taping, ankb brace and sion ROM (EROM), DROM, and PROM. All ROM data were converted to degrees if they were originally range of motion, and ankle support. Furthermore, adreported in radians. These ROM values were prelimiditional data were obtained from cross-referencing nary dependent measures, because they were later other published studies. A total of 82 titles were found in our literature search. Of these studies, each used to create the effect sizes analyzed in the statistihad to meet these criteria to be included in the anal- cal model. The effects of 2 independent variables were anaysis: (1) nonimpaired subjects were used in all studlyzed in this study. The first treatment variable was ies, (2) means and SDs for all data needed to be resupport condition with 4 levels: control (coded as 0), ported, (3) all studies were published in the English language, (4) nontreatment control conditions need- tape (coded as l ) , lace-up (coded as 2), and semiriged to be included, and (5) studies were published in id (coded as 3). The control condition was defined as measurements taken on a group with no type of peer-reviewed journals. Additionally, when determin-
Literature Search
J Orthop Sports Phys Ther.Volume SO. Number 4.April2000
,.- -
--
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
TABLE 1. Characteristicsof the studies used in the analysis.
Study
No./Sex*
Age (y)t
Mass (kg)t
Height (an))
InstrumentS
Support§
Time11
DeClercq5 Gehlsen et at9 Greene and Wightt4 Greene and Hillman12 Greene and Roland" Gross et all6 Gross et all5 Gross et all8 Gross et all7 Johnsonet aIz0 Kimura et al" Lindley and Kernozekx Martin and Hatterz3 Metcalfe et alZ4 Myburgh et atz6 Paris et aI2' Pederson et aP8 Seitz and Goldfuss" Wi lkersonJ6
7M 10 M 24 F 14 F 15 M, 15 F 8M,8F 8 M, 8 F 8 M, 8 F 2M,9F 13M,5F 8M, 10F 11 M 5 M, 5 F 10 F 12 M 30 M 15 M 29 M 30 M
23.0 2 1.3 23.5 2 3.7
75.5 2 3.6 78.1 2 10.0
184.0 2 0.1 177.7 2 4.9
-
-
-
-
72210.6 65.5 2 6.3 62.0 2 4.4
174.525.5 172.8 2 2.4 169.7 2 6.4
V CON ELGON ELGON ELGON ELGON ELGON ELGON ELGON ELGON ELGON V V CON ELGON ELGON V CON GON
0, 2 0,1,2,3 0, 2, 3 0, 1,3 0, 3 O,1,3 0, 3 0, 1, 2, 3 0, 1,3 0, 3 0, 3 0, 1,3 0,1,2,3 0, 1,2 0, 1 0, 1,2 0, 1 0, 1 0, 1
N N Y Y Y Y Y Y Y N N N Y Y Y Y Y Y Y
-
24.524.6 21.7 ?r 4.1 26.0 ?r 4.3
-
-
-
-
-
21.121.7 23.4 2 2.5 26.5 2 3.7
91.5212.8 67.7 2 9.5 57.2 2 7.7
186.927.6 171.3 2 7.1 166.0 2 6.0
22.0 2 3.3 22.9 2 3.3
83.4 2 10.5
180.1 2 6.1
-
-
-
-
-
-
-
-
Support assien- Ranee of menvtl rno~0110 C C R R C C C C C C C C C C C C C
c C
P, D, I, E P,D I,E I,E I,E I,E I,E I,E I,E I I P,D I P, D, I, E P, D, I, E P, D, I, E I P,I P,I
M indicates male; F, female.
t Mean 2 SD. # V indicates video system; CON, goniometer; and ELGON, electrogoniometer. See text for code descriptions. 11 Y indicates measured a time effect; N, did not measure a time effect. 7 R indicates random assignment; C, counterbalanced design. # P indicates plantar flexion; D, dorsiflexion; I, inversion; and E, eversion.
support. Cases classified in the tape condition included all conditions that provided support using adhesive and elastic athletic tape applied in a general basket-weave formation.' The lace-up condition included all cases that provided support using a soft canvaslike or nylon material2"."' (ie, McDavid, Swede-0, Kallasay). The semirigid condition included all cases where the support was offered through a firm thermoplastic material comprising a stirrup or posterior rigid supportL5." (ie, Aircast Sport-Stirrup, Active Ankle, DonJoy Ankle Ligament Protector). It was assumed in this analysis that different types of lace-up braces and semirigid braces within each group have similar mechanical properties. The second independent variable analyzed in this study was time. Specifically, all cases where the effects of brace support were evaluated before exercise (coded as 0) and after exercise (coded as 1) were used in the analysis. Because the studies that compared support over time did not all document the length of the time during exercise under each brace condition, we were not able to precisely code our exercise effects into various increments of time (ie, 20 minutes after, 40 minutes after, 60 minutes after).
Calculation of Effect Size and Statistical Analysis
sent the average difference between the experimental and control group divided by the pooled SDs of the groups. A standardized effect size of 1.0 would indicate that the average subject in the experimental group performed 34 percentile units higher than the control group subjects on the given outcome measure. Cohen4 provided general guidelines for effect size interpretation: an effect size of approximately d = 0.2 would be considered small, d = 0.5 would be considered medium, and d = 0.8 would be considered large. However, effect sizes should be evaluated considering the outcome variables they describe. Although Cohen's4 recommendation for effect sizes is based on social science research, based on the literature surveyed, we consider an effect size of d = 0.5 to be small, d = 1.0 to be medium, and d = 1.5 to be large for our study. The dependent variables used for this analysis were the standardized effect sizes for IROM, EROM, DROM, and PROM. The studies collected for this meta-analysis used repeated-measures designs, where the subjects' ROMs were measured under control and support conditions. These effects may have also been measured before and after exercise. The effect sizes for this study were computed as measures of a standardized mean change using the procedures outlined by Beckergdefined as follows:
An effect size is a standardized measure of change associated with treat~nent.~ This meta-analysis used the standardized measures of effect sizes that repre172
J Orthop Sports P h p Ther*Volume SOaNumber 4oApril 2000
TABLE 2. Range of motion (R t SD, degrees) used to calculate standardized effect sizes by support and time. Suppofl
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
Time
Control
Pre-exercise Inversion Eversion Dorsiflexion Plantar flexion
48.3 43.4 27.8 47.8
2 6.8
Postexercise Inversion Eversion Doniflexion Plantar flexion
34.6 17.2 44.3 46.7
2 6.3
t 8.3 t 6.7 t 6.1
t 4.0 t 5.2 t 4.9
Tape
27.5 27.1 33.6 40.4
Lace-up
2 6.4
t 6.5 t 5.6 2 5.2
36.1 t 7.0 29.0 6.1 30.6 t 6.1 37.5 t 4.6
+
Semirigid
+
30.2 2 26.4 t 15.8 t 27.8 t
5.6 6.0 5.8 6.1
32.5 5.8 35.7 t 4.7 49.2 t 4.1 74.8 t 6.1
41.5 33.0 27.0 33.4
5.3 5.4 8.0 6.2
38.1 t 4.9 38.0 t 5.8 NA* NA
t t t t
* NA indicates that no data were reported for semirigid brace after exercise for dorsiflexion and plantar flexion.
where d is the effect size that corresponds to the difference between the average scores of the experimental and control (Y)conditions divided by the control condition SD of sample j in study i. An effect size was calculated for each support condition in study i at pretest (before exercise) and posttest (after exercise), with the no-support condition in study i used as the control. To correct for small sample sizes, the effect sizes, d, were multiplied by the correction factor (1 - 3/(4n - 5)), where n is the sample size. This procedure was used to make d unbiased from being overestimated.'"he negative effect sizes obtained and subsequently used in the statistical analysis represent a reduction in ROM compared with the control condition. This was expected, since the support conditions evaluated in this study demonstrated some level of restriction compared with no support (control condition). A positive effect size would have suggested that the control condition offered greater ROM restriction compared with the support conditions. To test for differences between types of brace (tape, lace-up, and semirigid) and time of measurement (pre-exercise and postexercise) and the interaction between these factors, a mixed-model factorial
(a
TABLE 3. Standardized effect sizes (t movement.* Time
+ SE),
analysis of variance was applied to the effect size data. Studies was considered a random factor, whereas testing time, support condition, and the interaction between time and support were considered fixed factors. This method of analysis was selected to account for the dependencies within studies of time and type of support. These dependencies were present because the collected studies used exclusively repeated-measures designs with each group exposed to all conditions (ie, within-subjects designs). Simple main effects testing and the Tukey multiple comparison procedure were used to locate specific groups for each significant dependent measure. The level of probability for statistical significance was established a priori at P 5 .05.
RESULTS Descriptive information concerning the studies used is provided in Table 1. Table 2 illustrates the ROM values for each support condition and time that were used to calculate the effect sizes. The standardized effect sizes, number of cases associated with support and time, and restrictions on IROM, EROM, DROM, and PROM are presented in Table 3. In es-
number of cases (n),and average range of motion (degrees) restriction for support and time for each
Taw
Suppofl Lace-up
Semirigid
lnversion Eversion Dorsiflexion Plantar flexion Postexercise lnversion Eversion Dorsiflexion Plantar flexion
* Negative effect sizes indicate greater restriction in range of motion for each brace compared with the control condition. NA indicates that no data were reported for semirigid brace after exercise; thus, an effect size could not be created for dorsiflexion and plantar flexion. J Orthop Sports P h p Ther-Volume 30 Number 4 .April 2000
173
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
Pre
Post
Tape
Pre
Post
Lace-Up
Pre
Post
Semi-Rigid
Pre
Post
Pre
Tape
Post
Pre
Lace-Up
Post
Semi-Rigid
FIGURE 1. Standardized effect size for inversion range of motion by support and time. Values represent i 2 SE.
FIGURE 2. Standardized effect size for eversion range of motion by support and time. Values represent i 2 SE.
tablishing the standardized effect sizes, each brace condition significantly differed from the control condition (no brace). Specifically, for IROM an interaction was found between support and time (E,,, = 3.27, P = .04). Simple main effects testing revealed that at time 0 (before exercise) and at time 1 (after exercise), the semirigid support was more restrictive than lace-up (P< .05) and tape ( P < .05) (Figure 1). However, no difference was found between the tape and lace-up conditions at both levels of time (P > .05). In evaluating the effects of each condition across time, tape and lace-up support provided less restriction after exercise (P < .05) (Figure 1). The semirigid support, however, was still able to maintain adequate restriction following exercise ( P > .05). With respect to EROM, significant main effects were found for support (&, = 11.8, P < .001) and time (F,,, = 10.5, P = .002), although no interaction was noted (E,,, = 1.6, P = .214) (Figure 2). Post hoc testing revealed that the semirigid support was more restrictive than the lace-up (P< .05) and tape (P< .05) conditions (Figure 2). The lace-up condition also provided a greater reduction in eversion compared with the tape condition (P< .05). When collapsed across all levels of support, postexercise EROM values became more positive (P< .05), thus indicating a decrease in restriction compared with pre-exercise (Figure 2). For both DROM and PROM, only the tape and lace-up conditions were analyzed for the independent variables support and time. The semirigid brace condition was removed from these 2 models because effect sizes could not be created for semirigid postexercise scores (no data found in literature). Furthermore, semirigid orthoses are not mechanically designed to restrict these movements; thus, studying these effects did not appear warranted. When analyzing the effects of support and time on DROM, a main effect was found for support (F,,, = 5.71, P =
.026) (Figure 3). The tape condition was more restrictive than the lace-up condition ( P < .05). As for the main effect of time, no difference existed (Fl,, = 3.42, P = .078; 1 - P = .43). No interaction existed between both treatment variables as well (F,., = 1.98, P = .173). For PROM, only the main effect for time was significant (FIX,= 13.54, P = .001), where it was observed that overall restriction lessened after exercise (Figure 4). No support main effect existed = .276, P = .604; 1 - P = on PROM as well (F,,nn .lo).
174
DISCUSSION The primary mechanical function of ankle support is to restrict frontal plane motion of the foot and ankle joints. Particularly, it is the inversion or lateral ankle sprain that is most prevalent in sports and physical activities. With respect to this meta-analysis,
Pre
Post
Tape
Pre
Post
Lace-Up
FIGURE 3. Standardized effect size for dorsiflexion range of motion by support and time. Values represent i2 SE. J Orthop Sports Phys Ther.Volwne SO.Nurnber 4.April2000
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
by lace-up and tape support early after exercise activity. Both the tape and lace-up conditions decreased restriction after exercise, a change in ROM of a p proximately 9" (Table 3). Since both supports are composed of softer materials, this result is not surprising. It has been demonstrated that the longer the duration of exercise, the less restriction provided by Because the effect of support was not evaluated at specific durations of exercise in this study, it difficult to accurately determine how ROM changes at certain time intervals. However, it has been documented that tape loses its restrictive properties within 10 minutes after e x e r c i ~ eBecause .~ most exercise sessions of the studies used in this analysis lasted Pre Po3 Pre Posl more than 10 minutes and involved performing variTape Lace-Up ous agility and running activities, it can be deduced FIGURE 4. Standardized effect size for plantar flexion range of motion by that moderate intensity exercise greater than 10 minsupport and time. Values represent ic 5 SE. utes is sufficient enough to cause weakening of softer support (tape and lace-up braces), leading to less joint restriction. The primary importance regarding the most important assessment involved understandour results is deciding which type of support to use ing the effects of support and time on IROM. The and when. If ankle support is recommended for the results showed that the semirigid condition demonnonimpaired ankle as a preventive measure, tape and strated greater restriction compared with the tape lace-up braces do offer considerable restriction ( a p and lace-up conditions before and after activity. proximately 12 and 13%, respectively) compared When comparing this effect across time, the semirigwith no support. From a mechanical perspective, it is id condition demonstrated a restriction of approxiour contention that these methods of support offer a mately 23" compared with the control condition, good compromise between restriction and mobility. whereas the tape and lace-up conditions restricted inOur data suggest that both lace-up braces and adheversion by approximately 12 and 13" compared with sive tape support are equally effective. On the other the control condition, respectively (Table 3). Essenhand, since semirigid braces offer the greatest s u p tially, restriction offered by the semirigid support port before and after exercise, we advocate their use represents nearly a 45% improvement over the other for the athlete who has a history of ankle injuries or conditions for IROM. The fact that the semirigid who has chronic ankle sprains. condition provided more restriction than the other Injuries that occur from an excessive eversion mosupport conditions before activity is indicative of the tion are not as common compared with inversion stronger material used to prevent this frontal plane sprains, although such injuries are often more semotion. Perhaps even more important is that the vere.' Similar to the results found for IROM, very semirigid condition maintained its restrictive support, large effects were found among the support condiwhereas the tape and lace-up conditions decreased tions for eversion (Table 3). Specifically, the semirigsupport over time. These results clearly indicate that id support condition provided greater restriction with activity the effectiveness of the softer protective compared with both the lace-up and tape conditions. devices is reduced, whereas the more rigid material These results are not surprising considering the thermaintains its full restrictive property. These findings moplastic composition of the devices in restricting are not surprising given the abundance of previously frontal plane movement of the foot. Contrary to reported ~ t u d i e s ~ that ~-~ have ~ ' shown ~ . ~ continued what was found for IROM, the lace-up brace providinversion restriction with a semirigid device following ed greater eversion restriction compared with the a p exercise. Furthermore, the amount of restriction p r e plication of adhesive tape. This finding may be due vided by the semirigid condition for inversion was to the fact that so much emphasis is placed on rethe largest compared with the other movements. stricting inversion during application; however, when No differences were found between the lace-up applying the lace-up brace, it is essentially pulled and tape support conditions before and after exeronto the ankle and foot and fastened with laces or cise for IROM, although both support devices had Velcro fasteners. Furthermore, there is no emphasis diminished restriction after exercise. Although indito restrict IROM with the lace-up support applicavidual studies have shown differences between these tion. For EROM, a significant time effect was found, 2 ~onditions,'~:~~:'~ our overall findings were not in where overall restriction decreased after exercise. agreement. The results of our study may best be explained by the similar mechanical stability provided The softer, less rigid material used in the construcJ Orthop Sports Phys Ther.Volrme 30. Number 4.April 2000
17.i
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
tion of lace-up braces and adhesive tape is the main factor underlying this finding. When interpreting the results for PROM, it was found that both the lace-up and tape conditions displayed similar restrictive properties (Table 3). This finding can be attributed to the fact that both methods are designed to prevent plantar flexion movement. Structurally, the lace-up brace is able to prevent plantar flexion because of the material supporting the anterior portion of the ankle and the midfoot region. The purpose of the basket-weave tape application in preventing plantar flexion can be explained by its application technique.' Essentially, plantar flexion restriction is offered by dorsiflexing the ankle as stirrups and half anchors are applied.' No time effect or support by time interaction was found for PROM. This provides strong evidence that each of the appliances function similarly in restricting plantar flexion before and after exercise. One of the main criticisms concerning the use of adhesive tape is that it loosens up shortly after exercise.FJ However, based on this comprehensive statistical analysis of the literature, it does not appear that this occurred for plantar flexion. The practical application of this result is that both methods of support emerge as being equally effective. Perhaps the decision of whether to use either prophylaxis should be based on individual athlete preference and costefficiency rather than mechanical stability. For DROM, the tape condition provided more restriction than the lace-up condition when colIapsed across time. This result is primarily due to the large restriction of approximately 1lo offered by the tape condition for the pretest. With respect to mechanical support, application of adhesive tape may provide a stiffer environment around the foot and ankle compared with the softer lace-up brace. As stated earlier in this discussion, the primary mechanism of injury involving ankle sprains is forced inversion of the foot while the ankle is plantar flexed. Thus, the support offered by ankle braces and adhesive tape is not intended to restrict ankle dorsiflexion. So although the tape condition demonstrated greater dorsiflexion restriction than the lace-up brace, the clinical importance of this result is not readily apparent.
CONCLUSIONS This study provided a standardized, comprehensive statistical evaluation of the existing literature concerning ankle support and exercise on ankle and foot joint ROM. Our analysis showed that the greatest amount of inversion restriction was offered by semirigid braces compared with lace-up and tape conditions before and after exercise. The semirigid condition also provided the greatest restraint compared with the other support conditions. Furthermore, the lace-up condition also demonstrated more
restriction compared with the tape support. In evaluating sagittal plane motion, the adhesive tape condition offered superior ability in limiting ankle dorsiflexion compared with the lace-up condition; however, with respect to PROM, both the lace-up and tape condition were equally effective. Lastly, DROM remained unaffected following exercise, whereas PROM increased.
REFERENCES 1. Arnheim DD, Prentice WE. Principles ofAthletic Training. 8th ed. St Louis, Mo: Mosby Year Book Publishing; 1993. 2. Bahr R, Karlsen R, Lian 0, et al. Incidence and mechanisms of acute ankle inversion injuries in volleyball: a retrospective cohort study. Am ) Sports Med. 1994;22: 595-600. 3. Becker B. Synthesizing standardized mean-change measures. Br l Math Stat Psychol. l 988;41:257-278. 4. Cohen J. Statistical Power Analysis for the Behavioral Sciences. Rev. Ed. New York, NY: Academic Press; 1977. 5. DeClercq DLR. Ankle bracing in running: the effect of a push type medium ankle brace upon movement of the foot and ankle during the stance phase. lnt) Sports Med. 1997;18:222-228. 6. Garrick JG.The frequency of injury, mechanism of injury, and epidemiology of ankle sprains. Am ) Sports Med. 1977;5:241-242. 7. Garrick JG. Epidemiologic perspective. Clin Sports Med. 1982;1:13-18. 8. Garrick JG, Requa RK. The epidemiology of foot and ankle injuries in sports. Clin Sports Med. 1988;7:29-36. 9. Gehlsen GM, Pearson D, Bahamonde R. Ankle joint strength, total work and ROM: comparison between prophylactic devices. Athl Train JNATA. 1991;26:62-65. 10. Glass GV, McGaw B, Smith ML. Meta-analysis in Social Research. Beverly Hills, Calif: Sage Publications; 1981: 117. 11. Click JM, Gordon RB, Nishimoto D. The prevention and treatment of ankle injuries. Am ] Sports Med. 1976;4: 136-141. 12. Greene TA, Hillman SK. Comparison of support provided by a semirigid orthosis and adhesive ankle taping before, during and after exercise. Am ) Sports Med. 1990;18: 498-506. 13. Greene TA, Roland GC. A comparative isokinetic evaluation of a functional ankle orthosis on talocalcaneal function. ] Orthop Sports Phys Ther. 1989;11:245-252. 14. Greene TA, Wight CR. A comparative support evaluation of three ankle orthoses before, during and after exercise. I Orthop Sports Phys Ther. 1990;11:453-466. 15. Gross MT, Ballard CL, Mean HG, et al. Comparison of DonJoy ligament protector and Aircast Sport Stirrup orthoses in restricting foot and ankle motion before and after exercise. ) Orthop Sports Phys Ther. 1992;16:60-67. 16. Gross MT, Batten AM, Lamm AL, et al. Comparison of donjoy ankle ligament protector and subtalar sling ankle taping in restricting foot and ankle motion before and after exercise. ) Orthop Sport Phys Ther. 1994;19:33-41. 17. Gross MT, Bradshaw MK, Ventry LC, et al. Comparison of support provided by ankle taping and semirigid orthosis. J Orthop Sports Phys Ther. 1987;9:33-39. 18. Gross MT, Lapp AK, Davis JM. Comparison of swede-ouniversal ankle support and aircast sport-stirrup orthoses and ankle tape in restricting eversion-inversion before and after exercise. I Orthop Sports Phys Ther. 1991;13: 11-19. J Orthop Sports Phys Ther.Volume 30.Number 4.April 2000
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
Hedges LV, Olkin I. Statistical Methods for Meta-analysis. spatting and ankle taping on inversion before and after New York, NY: Academic Press; 1985:147-165. exercise. J Athl Train. 1997;32:29-33. Johnson RE, Veale JR, McCarthy GJ. Comparative study of 29. Rarick GL, Bigley G, Karst R, Malina RM. The measurable ankle support devices. J Am Podiatr Assoc. 1994;84:107support of the ankle joint by conventional methods of taping. J Bone loint Surg Am. 1962;44:1183-1191. 114. Kimura IF, Nawoczenski DA, Epler M, Owen MG. Effect 30. Rovere GD, Clarke TJ, Yates CS, et al. Retrospective comof the airstirrup in controlling ankle inversion stress. ] Orparison of taping and ankle stabilizers in preventing ankle thop Sports Phys Ther. 1987;9:190-193. injuries. Am ISports Med. 1988;16:228-232. Lindley TR, Kernozek TW. Taping and semirigid bracing 31. Seitz CJ, Goldfuss AJ. The effect of taping and exercise on may not affect ankle function range of motion. ] Athl passive foot inversion and ankle plantarflexion. Athl Train JNATA. 1 984;1 9:1 78-1 82. Train. 1995;3O:lO9-112. Martin N, Harter RA. Comparison of inversion restraint 32. Sitler MR, Horodyski M. Effectiveness of prophylactic anprovided by ankle prophylactic devices before and after kle stabilisers for prevention of ankle injuries. Sports Med. exercise. IAthl Train. 1993;28:324-329. 1995;20:53-57. Metcalfe RC, Schlabach GA, Looney MA, et al. A com- 33. Sitler M, Ryan J, Wheeler B, et al. The efficacy of a semirigid ankle stabilizer to reduce acute ankle injuries in basparison of moleskin tape, linen tape, and lace-up brace ketball: a randomized clinical study at West Point. Am ] on joint restriction and movement. J Athl Train. 1997;32: 136-140. Sports Med. 1994;22:454-461. Miller EA, Hergenroeder AC. Prophylactic ankle bracing. 34. S u ~ Ie, Schwellnus MP, Noakes T, et al. A fivefold reducPediatr Clin N Am. 1990;37:1175-1185. tion in the incidence of recurrent ankle sprains in soccer Myburgh KH, Vaughan CL, lsaacs S. The effects of ankle players using the sport-stirrup orthosis. Am J Sports Med. guards and taping on joint motion before, during, and 1994;22:601-606. after a squash match. Am J Sports Med. 1984;12:441- 35. Tropp H, Askling C, Gillquist J. Prevention of ankle 446. sprains. Am J Sports Med. 1 985;1 3:259-262. Rris DL, Vardaxis V, Kokkaliaris J. Ankle ranges of motion 36. Wilkerson GB. Comparative biomechanical effects of the during extended activity periods while taped and braced. standard method of ankle taping and a taping method J Athl Train. 1995;30:223-238. designed to enhance subtalar stability. Am J Sportr Med. Pedenon TS, Ricard MD, Merrill G, et al. The effects of 1991;19:58&595.
J Orthop Sports Phys Ther-Volume 3O.Number 4.April 2000
177
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
Invited Commentary Glass and colleague^'^ originally described metaanalysis as "the statistical analysis of a large collection of analysis results from individual studies for the purpose of integrating the findings. It connotes a rigorous alternative to the casual, narrative discussion of research studies which typify our attempts to make sense of the rapidly expanding research literature."'"fl Meta-analysis has evolved over the past 20 years to become an essential tool for integrating and interpreting research in the social, behavioral, and clinical science^.^ The evolution has not been without controversy.' There is no question, however, that meta-analysis has contributed significant new findings to multiple fields and that it is here to stay.4 This commentary is a reflection on the use of meta-analysis by Cordova, Ingersoll, and LeBlanc and an outline of some recent developments in quantitative synthesis. Cordova and colleagues%ave used meta-analysis to examine the effect of different types of ankle s u p port on range of motion before and after activity. They state that "the purpose of the study was to use a meta-analytic approach to evaluate ankle support and foot range of motion before and after exercise." Cordova et a1 conclude that the results of the investigation should "help clinicians make rational decisions concerning the selection of ankle appliances for preventing acute or chronic reinj~ry."*(p~~~) Using existing research to help guide clinical decision making is an important goal of research synthesis. A primary purpose of meta-analysis is to use procedures of statistical aggregation to achieve a consensus that informs clinicians, administrators, or policy makers.IR This commentary is divided into 5 sections, based on the stages involved in conducting a quantitative synthesis (meta-analysis) as described by C o ~ p e r . ~ The stages identified by Cooper are: Problem Formation, The Literature Search, Data Evaluation, Analysis and Interpretation, and Public Presentation. R o b l a Formation. Problem formation involves defining the parameters of the meta-analysis and the variables involved. The techniques of meta-analysis can be used for a variety of reasons. Originally, the approach was used to help interpret large numbers of primary studies that appeared to have conflicting statistical results. Establishing a degree of statistical consensus in a body of conflicting studies is still one of the primary functions of meta-analyses. Cordova et a1 discuss some other advantages of statistically aggregating studies such as improved statistical power and the ability to generalize findings across a broader population. Modern meta-analysis techniques also
provide a way to examine potential moderator variables and complex treatment by subject interactions if the sample of studies is large enough." In their justification for conducting a meta-analysis on the research examining the effectiveness of ankle supports, Cordova et a1 note that "the results of comparative studies are not all in agreement." They state that this is due to a variety of factors such as different techniques used to measure ROM, different applications of adhesive tape, different styles of braces, inadequate research design, and lack of statistical power (small sample sizes). These are all potential moderator variables. The authors make a reasonable case for using meta-analysis for the research question that they are interested in studying. They note that the general effectiveness of ankle bracing was discussed in 2 recent reviews and that the reviews "relied on a narrative approach, where outcomes of various studies were merely described and not statistically analy~ed."~(p"~) Just because a review is done in the traditional narrative format, however, does not mean the results are clinically or theoretically compr~mised.~ What did the 2 narrative reviews conclude? If one of the reviews concluded that ankle bracing is effective and the other review concluded that ankle bracing is ineffective, then there is a strong justification for using meta-analysis to resolve the confusion (see RedingIg and DobkinlVor a good example of conflicting narrative reviews). The Literature Search. When conducting a metaanalysis, individual studies are the sampling unit. The researcher must describe how the studies were selected and identify the inclusion or exclusion riter ria.^ Unlike a primary study, a researcher conducting a meta-analysis may attempt to include all studies in the "population." This involves identifying published articles, conference presentations, theses, dissertations, and other studies that have not been published. Unpublished or studies that are difficult to locate are referred to as "fugitive" literature and procedures have been developed to identify and retrieve investigations from the fugitive l i t e r a t ~ r e . ~ Cordova et a1 limited their search to studies that were published in peer-reviewed journals between 1966 and 1997. They examined 2 primary databases and provide a listing of the keywords used in the search. Additional titles were discovered in the references of the published studies. Cordova et a1 list the criteria for an investigation to be included in their meta-analysis. The section on the literature search J Orthop Sports Phys Ther.Volume 30-Number 4.April2000
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
strategies would have been more informative if they justified why they decided to focus only on studies appearing in the peer-reviewed literature. A question frequently used in evaluating a meta-analysis is: How likely is it that important, relevant studies were m i ~ s e d ?Our ~ . ~ confidence in the results of a review is greater when we are certain that no relevant and highquality studies, either published or unpublished, were missed. Without some explanation and justification for the search strategy that was used by Cordova et al, the question of whether relevant highquality studies were missed remains open. Cordova et a1 note that they were not blind to the findings in each study when decisions were made to reject or accept an article for inclusion in the metaanalysis. To avoid bias, they state that any study that failed to meet all of the inclusion criteria was not included in the analysis. While this information is help ful, it does not answer all the questions a reader might have about why a particular study was included or excluded. No information is provided on how the studies were reviewed for inclusion. That is, were all studies examined by a single rater who made decisions regarding each of the inclusion criteria? If more than one person reviewed studies for possible inclusion in the meta-analysis, what was the degree of interrater agreement across the inclusion criteria? Research over the past several years has suggested that a bias may exist in selecting ~tudies.~ To reduce this bias, Chalmers et al have developed procedures that include reviewing the article without author attribution or affiliati~n.~.~ Often, the results section of the article is removed prior to examination so that the reviewer is making a decision on whether or not to include the study based on the design and methods rather than the r e s ~ l t s . " ~ ~ . ~ ~ Data Evaluation. The data evaluation stage of a meta-analysis involves making judgements about whether or not individual data points are trustworthy enough to be included in the research. Data evaluation requires that the researcher identify those elements from the primary studies that are relevant to the research question and establish criteria for judging the adequacy of the procedures used to gather and code the data. Cordova et a1 selected and coded a number of demographic factors along with data on type of ankle support and changes in range of motion. Information was also collected regarding the sample size and results of statistical comparisons. Just as researchers who are conducting primary studies must be concerned with the accuracy and reliability of data collected from individual subjects, a researcher who is conducting a meta-analysis must be concerned with the accuracy and reliability of information obtained from the primary s t u d i e ~Cordova .~ et a1 do not provide details regarding who collected information from the studies or how it was coded. While previous J Orthop Sports Php Ther*Volume SO. Number 4.April 2000
research on the reliability of coding information from primary studie~l'.~~ suggests that the accuracy of coding and the agreement between raters is generally high, the reliability and accuracy for coding information about variables in a particular meta-analysis must be examined and reported by the investigators. No information is included on the reliability of judgements made by raters in the Cordova et a1 investigation. Of particular note is the lack of information on the consistency of computing the effect size values. For a detailed discussion of issues associated with study coding and reliability, see and Orwin.I7 An important moderator variable, absent from the list of coded variables in the Cordova et a1 investigation, is research quality (type of design). Cooper and Hedges7 state that the aspect of coding studies for meta-analysis that "engenders the most discussion" is how to represent differences in research design. A variety of methods have been developed to code different designs and to weight research designs based on quality. Design quality variables, such as whether the outcome measure was blindly recorded, have been repeatedly shown to have an impact on effect size.21.24 Research design is viewed as an important moderator variable in the modern approach to metaanalysis and its absence in the Cordova et a1 investigation limits our ability to interpret the results. It is particularly puzzling given that the authors identify "inadequate research designs" as one of the factors contributing to the lack of agreement in the primary research studies evaluating the effectiveness of ankle supports. Data Analysis and Interpretation. During the analysis and interpretation stage, the separate data points collected by the researcher are synthesized into a unified statement about the research problem. This is generally accomplished by converting the statistical results of the aggregated studies into a common metric, referred to as an effect size. Cordova et al use the standardized mean difference (dindex) for a 2group comparison to assess differences in range of motion across the ankle support conditions. The & index is an appropriate effect size for the nature of the comparison that the authors are evaluating. The dindex was adjusted for sample size and then further analyzed using a mixed-model factorial analysis of variance (ANOVA). In the ANOVA, studies were considered a random factor and testing-time, support condition, and the interaction between time and support were considered fixed factors. The authors note that this model was selected to account for possible dependencies among effect sizes. That is, more than one effect size (&index) was generated for some studies. The problem of dependency among multiple effect sizes generated within a single study is an issue that has been debated since metaanalysis was introduced by Glass et a1 in the mid1 9 7 0 ~ . ' "The ~ ~ approach taken by Cordova et a1 to 179
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
analyze the data (ie, ANOVA) was used widely in the 1980s, but is not currently advocated for a variety of r e a ~ o n sCooper5 .~ recently observed the following: Standard inferential procedures were the techniques used initially by meta-analysts for examining variance in effects. Glass, McCaw, and Smith (1981) detailed how this approach is carried out. At least two problems arise with the use of traditional inference procedures in research synthesis, however. The first is that traditional inference procedures do not test the hypothesis that the variability in effect sizes is due solely to sampling error. . . . Also, because effect sizes can be based on different numbers of data points, they can have different sampling variances associated with them-that is, they are measured with different amounts of errorYw)
The first step in most modern meta-analyses involves conducting a test of homogeneity among effect sizes (eg, Q statistic)."I4 Homogeneity analysis compares the observed variance to that expected from sampling error. The homogeneity analysis addresses the following question: Is the observed variance in the collected effect sizes significantly different from that contributed by sampling error alone? If the answer is yes, then the researcher proceeds to examine study characteristics and variables, such as the type of ankle support, associated with variance in effect sizes. This analysis might include a form of regression (meta-regression) developed for use with effect sizes, a Bayesian analysis, or an expansion of the homogeneity analysis used in conjunction with confidence interwls.5.l "1" A small, but potentially important advancement in data reduction for meta-analysis relates to the adjustment of &indices. The approach used by Cordova et a1 to weight effect sizes by sample size is being replaced by adjusting the dindex by an alternative weighting factor (w,), which is the inverse of the variance associated with each dindex estimate. The advantage of using wi as the weight rather than the sample size is that precise confidence intervals can then be generated around the average effect size estimate.5 Public Presentation. The presentation of metaanalysis results generally follows the same format as for a primary research study. Cordova et a1 include appropriate sections describing the variables, methods, and analysis used in their investigation. Table 1 would have been more useful if the effect size values associated with each of the studies were reported. Table 3 includes the mean effect sizes and standard errors for the various ankle support devices and range of motion conditions, which are useful data. The information is then duplicated in Figure 1. Rather than duplicating the information from Table 3, Figure 1 would have improved the presentation if it included the individual effect sizes and their distribution (eg, a stem-and-leaf plot). Typical methods for presenting effect size values include stem-and-leaf plots or box plots. Specific methods for depicting the
relationship between effect size and sample size, such as funnel plots, are also an option.lR The discussion presents a clear overview and summary of the results, but I would have found it help ful if the meta-analysis findings were compared and contrasted to those of previous narrative reviews of the ankle support literature. The authors could have identified inconsistency in previous narrative reviews and described how those inconsistencies have been addressed by the quantitative synthesis. Since the analysis did not include moderator variables, there is no discussion of the potential impact of factors such as type of design, research quality, method of data collection, or sample characteristics on variability among effect sizes. For example, if studies where the outcome measures were blindly recorded also reported consistently larger (or smaller) effect size values, then recommendations could be made for the planning, design, and interpretation of future research." The points made above should be viewed as suggestions for future research, not as criticisms of the present investigation. Cordova, Ingersoll, and LeBlanc achieved their goal of providing information that may help clinicians make better decisions when selecting ankle supports. They synthesized the results of 19 studies that examine the effectiveness of different ankle supports and presented the statistical consensus and clinical implications of their research. New methods associated with meta-analysis, combined with advances in clinical treatment, will help ensure that future research efforts expand on the consensus and conclusions provided by Cordova et al. Kenneth J. Ottenbacher, PhD School of Allied Health Sciences University of Texas Medical Branch 301 University Blvd. Galveston, TX 775551028
[email protected]
REFERENCES Bangert-Drowns, RL. Review of developments in metaanalytic method. Psych Bull. 1986;99:388-399. Chalmers TC, Levin H, Sacks DS, et al. Meta-analysis of clinical trials as a scientific discipline: I control of bias and comparison with large co-operative trials. Stat Med. l987;6:315-328. Chalmers TC, Smith H Jr, Blackburn B, et al. A method for assessing the quality of a randomized control trial. Cont Clinic Trials. 1981;2:31-49. Cook TC, et al. Meta-analysis for Explanation: A Casebook. New York, NY Russell Sage; 1992. Cooper HM. Integrating Research: A Guide for Literature Reviews. 3rd ed. Newbury Park, Calif: Sage; 1998. Cooper HM, Hedges LV eds. The Handbook of Research Synthesis. New York, NY Russell Sage; 1994. Cooper HM, Hedges LV. Research synthesis as scientific enterprise. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY Russell Sage; 1994:3-14. J Orthop Sports Phys Ther.Volume 30.Number 4.April2000
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
Cordova ML, lngersoll CD, LeBlanc MJ. Influence of ankle support on joint range of motion before and after exercise: a meta-analysis. 1 Orthop Sports Phys Thec 2000;30: 170-1 77. Dear KB, Begg CB. An approach for assessing publication bias prior to performing a meta-analysis. Stat Sci. 1992; 7:237-245. Dobkin BH. Focused stroke rehabilitation programs do not improve outcome. Arch Neurol. 1989;46:701-703. Eagly AH, Wood W. Using research synthesis to plan future research. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sage; 1994:4485-4502. Eddy DM, Hasselblad V, Schacter R. Meta-analysis by the Confidence Interval Approach. San Diego, Calif: Academic Press; 1992. Glass GV, McGaw B, Smith ML. Meta-analysis In Social Research. Beverly Hills, Calif: Sage; 1981. Hedges LV. Statistical considerations. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sage; 1994:29-37. Hedges LV, Olkin I. Statistical Methods for Meta-analysis. Orlando, Fla: Academic Press; 1985. Louis TA, Zelterman D. Bayesian approaches to research synthesis. In: Cooper HM, Hedges LV, eds. The Handbook
17. 18. 19. 20. 21.
22.
23. 24.
of Research Synthesis. New York, NY: Russell Sage; 1994: 41 1-422. Orwin RG. Evaluating coding decisions. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sage; 1994:139-162. Petitti DB. Meta-analysis Decision Analysis and Cost-effectiveness Analysis. New York, NY: Oxford University Press; 1994. Reding MJ,McDowell FH. Focused stroke rehabilitation programs improve outcome. Arch Neurol. 1989;46:70&701. Rosenthal MC. The fugitive literature. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sage; 1994:85-94. Schulz KF, Chalmers TC, Hayes RJ,Altman DG. Empirical evidence of bias dimensions of methodological quality associated with estimates of treatment effects in controlled trials. IAMA. l995;273:408-4l2. Shadish WR, Haddock CK. Combining estimates of effect size. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sage; 1994: 261-281. Stock WA. Systematic coding for research synthesis. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sa e; 1994:125-138. Wortman PM. Judging research qJa7ity. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sage; 1994:97-109.
Author Response We thank Dr Ottenbacher for his overview of meta-analysis. We think that investigators who intend to perform a meta-analysis will benefit from his commentary. The framework of Dr Ottenbacher's commentary represents the primary stages in conducting a meta-analysis according to Cooper.' Dr Ottenbacher raises 7 issues concerning our work that warrants specific attention. The use of only peer-reviewed publications as the source of the sampling unit in our analysis is one such issue. When performing a meta-analysis, the data producing the effects should come from highquality, relevant studies. We determined that studies published in peer-reviewed journals that met our inclusion criteria would be considered highquality. There is no question that greater confidence in the results of a quantitative synthesis can be achieved if all of the relevant, high quality studies from the p o p ulation are i n ~ l u d e d . ' Perhaps .~ an argument could be made that some unpublished dissertations or theses could have contributed to our analysis; however, we felt that reporting effects from studies that have not undergone a peer-review process would jeopardize the quality of our analysis rather than enhance it. Because all of the studies used in our analysis tested the same hypothesis and were relevant to each other, it is unlikely that the nature and relationships of our J Orthop Sports Phys Ther-Volume 30. Number 4 .April 2000
findings are incorrect, although the magnitude of these effects might be overestimated.' A secondary issue raised by Dr Ottenbenbacher involving the methods of our literature search concerns the number of raters and the degree of rater agreement when determining which studies were included in the analysis. Undoubtedly, bias can be introduced when selecting studies to be included in a meta-analysis.' Our inclusion criteria were very clear and were not susceptible to differing interpretations. Nonimpaired subjects were used in all studies, means and standard deviations for all data needed to be reported, all studies were published in English, nontreatment control conditions needed to be included, and the studies that we analyzed were published in peer-reviewed journals. While inclusion criteria may differ among meta-analyses in general, there was very little opportunity for bias when choosing studies for our analysis. We concur with Dr Ottenbacher's comment that identifying who made the decisions to include the studies, along with the reliability of these decisions, would have been useful, but we did not consider this a limitation. Dr Ottenbacher raises 2 issues concerning the data evaluation stage of our meta-analysis. The first issue involves the reliability and accuracy of coding data from the studies that were used. As mentioned above, our inclusion criteria and the data that we
Journal of Orthopaedic & Sports Physical Therapy® Downloaded from www.jospt.org at Florida Gulf Coast University on November 17, 2017. For personal use only. No other uses without permission. Copyright © 2000 Journal of Orthopaedic & Sports Physical Therapy®. All rights reserved.
were looking to extract from these studies was very clear. For example, the data coded in Table 1 of our study shows that there was virtually no dispute in categorizing support conditions, or type of support within each of the studies. Therefore, the potential for making errors with respect to coding decisions was low, thus minimizing threats to reliability or validity. The second issue regarding our data evaluation procedures involves the coding of selected moderator variables, particularly type of design or research quality. Evaluating the impact or effect of research design can have large implications on effect sizes generated in the meta-analysis.%owever, in our analysis we coded the type of design used in each of the studies represented in our analysis. Because 17 of the 19 studies used similar repeated measures designs, we surmised that there was little dissimilarity between the studies. Given this, we disagree with Dr Ottenbacher that failure to consider the type of experimental design limits the interpretation of our data. If the studies that we included in our metaanalysis had more variability in research design, then we would agree with Dr Ottenbacher's suggestion that not considering moderator variables would limit interpretation of our data and results. With regard to the data analysis and interpretation stage of our analysis, Dr Ottenbacher commented on our use of the dindex (standardized mean difference for 2group comparisons) in estimating the effects sizes for our comparisons. Although it appears that Dr Ottenbacher agrees with the appropriateness and implementation of our use of the dindex within our mixed-model factorial analysis of variance, he raises 2 issues. These issues involve the adequacy of using standard inferential statistics when evaluating differences in effects sizes. He questions whether the variability in treatment effect. cannot be differentiated from variability due to sampling error, and that the effect sizes are computed with different amounts of error.' These are critical issues and assumptions that must be evaluated when applying inferential statistics to any data set. To evaluate the potential effects of variability due to sampling error, a test of homogeneity of effect sizes should be performed.' This procedure assesses whether or not the variance o b served with the treatment differs from the variance expected from sampling error. As mentioned in the commentary, the Qstatistic is primarily used to evalu-
ate the homogeneity of effect sizes.' In our study we initially tested the homogeneity of effect sizes using the Qstatistic, but did not think that this traditional method of assessment was appropriate for a mixedmodel because we had dependencies among our effect sizes. As far as the presentation of our meta-analysis, Dr Ottenbacher provides an array of comments pertaining to the illustration of our data within the tables and figures. Undoubtedly, we agree that other graphical illustrations (box plots or stem-and-leaf-plots) could have been used to portray our data. This information, however, was not requested by the reviewers at any point during the review process. Therefore, it is difficult to speculate whether these graphical illus trations would have added to the presentation of our results above and beyond the figures that were presented. Dr Ottenbacher's comments on our discussion concerning the integration of the narrative reviews cited4.Qompared to our quantitative synthesis are well-taken. In conclusion, we thank Dr Ottenbacher for his positive comments regarding the appropriateness of our article. Our work represents the only quantitative synthesis that exists regarding the efficacy of ankle bracing on limitingjoint range of motion before an after exercise. We agree with Dr Ottenbacher that we have achieved our goal of providing clinicians with definitive information to be used when making decisions concerning the application of external ankle support. Mitchell L. Cordova, PhD, ATC Christopher D. Ingersoll, PhD, ATC Michael J. LeBlance, PhD
REFERENCES Cooper HM. Integrating Research: A Guide for Literature Reviews. 3rd ed. Newbury Park, Calif: Russell Sage; 1998. Dear KB, Begg CB. An approach for assessing publication bias prior to performing a meta-analysis. Stat Sci. 1992;7: 237-245. Glass GV, McGaw B, Smith ML. Meta-analysis in Social Research. Beverly Hills, Calif: Sage; 1981. Miller EA, Hergenroeder AC. Prophylactic ankle bracing. Ped Clin North Am. 1990;37:1175-1185. Sitler MR, Horodyski M. Effectiveness of prophylactic ankle stabilisen for prevention of ankle injuries. Sports Med. 1995;20:53-57. Wortman PM. Judging research quality. In: Cooper HM, Hedges LV, eds. The Handbook of Research Synthesis. New York, NY: Russell Sage; 1994.
J Orthop Sports Phys Therevolume %).Number 4.April 2000