- Email: [email protected]
Recalibration and Validation of the Cumberland Ankle Instability Tool Cutoff Score for Individuals With Chronic Ankle Instability
Accepted Manuscript Recalibration and validation of the Cumberland Ankle Instability Tool cutoff score for individuals with Chronic Ankle Instability Cynthia J. Wright, PhD, ATC Brent L. Arnold, PhD, ATC Scott E Ross, PhD, ATC Shelley W Linens, PhD, ATC PII:
S0003-9993(14)00334-7
DOI:
10.1016/j.apmr.2014.04.017
Reference:
YAPMR 55822
To appear in:
ARCHIVES OF PHYSICAL MEDICINE AND REHABILITATION
Received Date: 26 November 2013 Revised Date:
14 April 2014
Accepted Date: 17 April 2014
Please cite this article as: Wright CJ, Arnold BL, Ross SE, Linens SW, Recalibration and validation of the Cumberland Ankle Instability Tool cutoff score for individuals with Chronic Ankle Instability, ARCHIVES OF PHYSICAL MEDICINE AND REHABILITATION (2014), doi: 10.1016/ j.apmr.2014.04.017. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Running Head: Recalibration of the CAIT in CAI individuals
Authors name and highest academic degree (in order of authorship) Cynthia J Wright*, PhD, ATC Brent L Arnold†, PhD, ATC Scott E Ross‡, PhD, ATC Shelley W Linens§, PhD, ATC
RI PT
Title: Recalibration and validation of the Cumberland Ankle Instability Tool cutoff score for individuals with Chronic Ankle Instability
M AN U
SC
Research was conducted in the Department of Health and Human Performance at Virginia Commonwealth University, Richmond, Virginia, USA. All authors were affiliated with Virginia Commonwealth University at the time of study completion. All authors have changed affiliations since study completion. Current affiliations are indicated below. * Whitworth University † Indiana University School of Health and Rehabilitation Sciences ‡ University of North Caroline Greensboro § Georgia State University
TE D
Portions of the content of this manuscript were presented at 2 national conferences: Wright CJ, Arnold BL, Ross SE, Pidcoe PE. (2011, June). Validation of a Recalibrated Cumberland Ankle Instability Tool Cutoff Score for Chronic Ankle Instability. Poster session for the 2011 National Athletic Trainer’s Association Annual Meeting and Clinical Symposium, New Orleans, LA. Published Abstract: J Athl Train 2011: 46(3)(Suppl):S124. Arnold BL, Wright CJ, Linens SW, Ross SE (2010, June). Recalibration of the CAIT cutoff score for chronic ankle instability. Poster presentation for the 2011 American College of Sports Medicine Annual Meeting, Denver, CO. Published Abstract: Med Sci Sports Exerc 2011: 43(5)(Suppl)S341.
EP
No financial support was received for this research. The authors have no conflicts of interest to disclose.
AC C
Shelley Linens reports a doctoral research grant from the National Athletic Trainers' Association Research and Education Foundation Corresponding author: Cynthia Wright Whitworth University 300 W Hawthorne Road Spokane, WA 99251 (509) 777-3244 Office (509) 777-4943 Fax [email protected]
ACCEPTED MANUSCRIPT
1
Recalibration and validation of the Cumberland Ankle Instability Tool cutoff score
2
for individuals with Chronic Ankle Instability
3
Objective: To independently recalibrate and re-validate the Cumberland Ankle Instability Tool
5
(CAIT) cutoff score for discriminating individuals with and without chronic ankle instability
6
(CAI). There are concerns the original cutoff score (≤27) may be suboptimal for use in the CAI
7
population. Design: Case-control. Setting: Research laboratory. Participants: Two independent
8
datasets were used. Dataset 1 included 61 individuals with a history of ≥1 ankle sprain and ≥2
9
episodes of giving-way in the past year (CAI), and 57 participants with no history of ankle sprain
M AN U
SC
RI PT
4
or instability in their lifetime (uninjured). Dataset 2 included 27 uninjured participants, 29 CAI
11
participants, and 26 individuals with a history of a single ankle sprain and no subsequent
12
instability (copers). Interventions: All participants completed the CAIT during a single session.
13
In Dataset 1 a receiver operating curve (ROC) was calculated using CAIT score and group
14
membership as the test variables. The ideal cutoff score was identified using Youden’s index.
15
The recalibrated cutoff score was validated in Dataset 2 using ROC analysis and clinimetric
16
characteristics. Main Outcome Measure(s): CAIT cutoff score and clinimetrics. Results: In
17
Dataset 1, the optimal cutoff score was ≤25, which is lower than previously reported. In Dataset
18
2 the recalibrated cutoff score demonstrated a sensitivity of 96.6%, specificity of 86.8%, positive
19
likelihood ratio (LR+) of 7.318, negative likelihood ratio (LR-) of 0.039. There were seven false
20
positives and one false negative. Conclusions: The recalibrated CAIT score demonstrated very
21
good clinimetric properties; all properties improved when compared to the original cutoff.
22
Clinicians using the CAIT should utilize the recalibrated cutoff score to maximize test
23
characteristics. Caution should be taken with copers, who had a high rate of false positives.
AC C
EP
TE D
10
1
ACCEPTED MANUSCRIPT
24 25
Keywords: Functional Ankle Instability; Ankle Sprain; Clinimetrics; Patient Questionnaire;
26
Self-Reported Measure; Test Characteristics
28
List of abbreviations:
29
CAI
30
CAIT Cumberland Ankle Instability Tool
31
ROC receiver operating curve
32
LR+
positive likelihood ratio
33
LR-
negative likelihood ratio
34
AUC area under the curve
RI PT
27
AC C
EP
TE D
M AN U
SC
Chronic Ankle Instability
2
ACCEPTED MANUSCRIPT
35
Chronic ankle instability (CAI) is a common sequelae of lateral ankle sprain, affecting approximately 32-47% of ankle sprain patients.1-3 A symptomatically defined pathology, CAI is
37
characterized by recurrent sprains and/or recurrent instability (e.g. episodes of “giving-way”)
38
after an ankle sprain.4, 5 Recent articles have reviewed the problems and variability involved with
39
a symptomatic definition of CAI.5, 6
40
RI PT
36
Patient questionnaires can serve various functions. One function can be to provide
reliable measurement of patient reported symptoms such as pain, functional limitations, and
42
instability occurrence with specific activities. A questionnaire that has been widely used in ankle
43
instability literature,7-10 translated into multiple languages,11, 12 and shown to be a significant
44
predictor of ankle instability status,13 is the Cumberland Ankle Instability Tool (CAIT).14 First
45
published by Hiller et al.14 in 2006, this nine question survey focuses on symptoms of instability
46
during several different physical tasks. The CAIT results in a score ranging from zero to 30 with
47
higher scores indicating higher stability. Original research established a cutoff score of ≤27 as
48
indicative of CAI group membership.14 We observed in our laboratory that the established cutoff
49
score appeared to be too high. Individuals who had a history of ankle sprain, but subjectively
50
reported that their ankle “didn’t really bother them” were occasionally classified as having CAI
51
based on the established cutoff score of ≤27. Perhaps because of this issue, some authors have
52
independently chosen to use a lower cutoff score (i.e. ≤23 and ≤24) than what was originally
53
validated by Hiller et al.15, 16 Recently, the International Ankle Consortium recommended that a
54
cutoff score of ≤24 be used in CAI inclusion criteria.17 However, to our knowledge statistical
55
evidence to support the selection of a lower cutoff value has not been reported in the literature.
56
Further investigation into the question of appropriate cutoff scores for the CAIT
57
AC C
EP
TE D
M AN U
SC
41
highlighted an important limitation in the criteria used to establish the original cutoff score.
3
ACCEPTED MANUSCRIPT
Hiller et al.14 used a history of ankle sprain alone to define group membership when calculating
59
the cutoff score. This created a group designation of “sprained” vs. “un-sprained”—yet the cutoff
60
score derived has been commonly used to define “CAI” vs. “No CAI” group membership. These
61
two categorization schemes have important differences, and we propose should not be used
62
interchangeably. Specifically, recent research has highlighted that some individuals (frequently
63
called “copers”) have a history of ankle sprain but no ongoing instability.7, 18-26 Copers score
64
similar to uninjured controls on questionnaires such as the Foot and Ankle Ability Measure
65
(FAAM), FAAM-Sport, and CAIT.7 Thus, inclusion of these individuals in the sprained group
66
could inflate scores and lead to the establishment of a higher cutoff value that would be
67
established if only symptomatic ankles were included.
SC
M AN U
68
RI PT
58
The purpose of the current investigation was to independently re-validate and potentially recalibrate a CAIT cutoff score, by including only individuals with a history of lateral ankle
70
sprain AND recurrent instability in the CAI group.13 We hypothesized that the resulting cutoff
71
score would be lower than previously reported. Additionally, we desired to test whether a
72
recalibrated cutoff score would appropriately classify subjects in an independent subject pool
73
including uninjured control subjects, copers and CAI individuals. We hypothesized that a lower
74
cutoff score would result in fewer false classifications of copers.
EP
AC C
75
TE D
69
76
METHODS
77
Subjects Dataset 1: Recalibration
78
Subjects were originally recruited for three independent research studies which collected
79
CAIT scores. Participants were recruited from a large metropolitan area, including a university
80
campus. The one hundred and eighteen individuals were recruited via direct contact with
4
ACCEPTED MANUSCRIPT
individuals and recruiting announcements in university courses and included in this study: 61
82
individuals with CAI and 57 uninjured individuals. Demographics are reported in Table 1.
83
Inclusion criteria for the CAI group included a history of at least one lateral ankle sprain >6
84
weeks prior to study entry and at least two reported episodes of giving way per year. Uninjured
85
individuals had no history of ankle sprain or instability in their lifetime. Exclusion criteria for
86
both groups included a history of lower extremity fracture or surgery, any acute symptoms of
87
ankle injury, or assisted ambulation. University Institutional Review Board approval was
88
obtained prior to data collection for both datasets.
89
Subjects Dataset 2: Validation
SC
M AN U
90
RI PT
81
Eighty-two subjects were recruited from a large metropolitan area, including a university campus via direct contact with individuals and recruiting announcements made in university
92
courses. Twenty-nine individuals with CAI, 26 copers and 27 uninjured individuals were
93
included in the dataset. Table 2 includes subject demographics. Inclusion criteria for the CAI
94
group and uninjured group were the same as in Dataset 1, with the addition that CAI subjects had
95
to be at least 1 year post-initial injury. Individuals categorized as copers reported a history of a
96
single lateral ankle sprain which required protected weight bearing, immobilization, and/or
97
limited activity for ≥ 24 hours, no perceived instability, and had resumed all pre-injury activities
98
without limitation for at least 12 months prior to testing. Perceived instability was assessed with
99
a single yes or no question, “Does your ankle ever give-way, roll-over or feel unstable?”
AC C
EP
TE D
91
100
Exclusion criteria were the same as in Dataset1. Additionally, subjects had to perform at least 90
101
minutes of physical activity per week; this activity could be of any intensity or mode. The CAIT
102
score was not used as an inclusion criteria for any group.
103
Data Collection
5
ACCEPTED MANUSCRIPT
104
In each study, subjects reported to the University Sports Medicine Research Laboratory and gave informed consent. Inclusion and exclusion criteria were verified, demographic data
106
(including age, height and weight) was collected, and the CAIT was completed. A custom
107
computer program (Access, Microsoft, Redmond, WA) recorded and scored CAIT questionnaire
108
responses for each subject. Only CAIT scores for the involved limb (CAI subjects) or
109
comparison limb (uninjured subjects) were utilized in the analysis. For CAI individuals with
110
bilateral instability, the most unstable ankle (i.e. lowest CAIT score) was included in the
111
analysis.
112
Statistical Analysis
SC
M AN U
113
RI PT
105
All analyses were performed using IBM SPSS Statistics 20 (Armonk, New York, USA). Paired t-tests were used to compare subject demographics between groups in Dataset 1, and a
115
one-way ANOVA was used to compare the same variables among groups in Dataset 2. For both
116
datasets, receiver operator characteristic (ROC) curves were generated with CAIT score as the
117
dependent variable, and group membership (CAI vs. no CAI) as the independent variable. In
118
Dataset 2, the no CAI group included both copers and uninjured subjects. Area under the curve
119
(AUC) was used to identify a significant ROC curve using a one-sided test (alpha = 0.05).
120
After calculating the ROC curve, the diagnostic sensitivity and specificity for each
EP
TE D
114
potential cutoff score were calculated. The largest Youden index value [sensitivity + specificity-
122
1)] was used to determine the ideal cutoff score.27 In Dataset 1, the clinical meaningfulness of
123
each cutoff score was evaluated by calculating the positive likelihood ratio (LR+) and negative
124
likelihood ratio (LR-). LR+ was calculated as [sensitivity/(1-specficity)], and LR- was calculated
125
as [(1-sensitivity)/specificity]. In Dataset 2, the purpose was to validate the recalibrated cutoff
126
score (≤25), thus clinimetric properties including the diagnostic sensitivity and specificity were
AC C
121
6
ACCEPTED MANUSCRIPT
calculated for the recalibrated cutoff score. Additionally, the clinical meaningfulness of the
128
recalibrated cutoff score was evaluated by calculating the LR+, LR-, number of false positives
129
and number of false negatives. A low LR- value (<0.2) would indicate that a negative test
130
substantially decreases the likelihood of an individual truly having CAI, a high LR+ (>5) would
131
indicate that a positive test substantially increases the likelihood of an individual truly having
132
CAI.28
SC
RI PT
127
134
RESULTS
135
Dataset 1: Recalibration
136
M AN U
133
The average CAIT score for the uninjured subject group was 29.53±1.04 (range, 26 to 30), and the CAI group was 19.41±4.27 (range, 8 to 28). Results of statistical tests on
138
demographic variables are reported in Table 1. The ROC was significant (AUC = 0.996,
139
p=0.005; Figure 1), indicating that CAIT score significantly predicted group membership. The
140
largest Youden index value (0.95) indicated that a CAIT score ≤25 was the ideal cutoff to
141
distinguish group membership (Table 3). High sensitivity (95.1%) and specificity (100%) were
142
calculated at this cutoff (Table 3). The LR- value was 0.049. Due to perfect specificity, a LR+
143
could not be calculated at the recalibrated cutoff score. However, the next nearest cutoff value
144
where LR+ could be calculated (26.5) resulted in a LR+ of 27.171.
145
Dataset 2: Validation
EP
AC C
146
TE D
137
The average CAIT scores by group were 28.93±1.69 (range, 23 to 30) for the uninjured
147
subjects, 27.31±2.02 (range, 23 to 30) for copers, and 19.59±4.15 (range, 6 to 26) for CAI
148
subjects. Results of statistical tests on demographic variables are reported in Table 2. The ROC
149
was significant (AUC = 0.988, p<0.001; Figure 2), indicating that CAIT score again significantly
7
ACCEPTED MANUSCRIPT
150
predicted group membership. The largest Youden index value (0.893) indicated that a CAIT
151
score of ≤23 was the ideal cutoff to distinguish group membership in this dataset (Table 4). The
152
Youden index value for the recalibrated cutoff score of ≤25 was only slightly lower (0.834). At the recalibrated cutoff score of ≤25, sensitivity (96.6%) and specificity (86.6%) were
RI PT
153
both high (Table 4 and Table 5). The LR- value was 0.039, and the LR+ was 7.318. Using the
155
recalibrated cutoff score there were seven false positives (one uninjured subject and six copers)
156
and one false negative (one CAI subject).
SC
154
157
159
DISCUSSION
M AN U
158
The purpose of the current investigation was to independently re-validate and potentially recalibrate a CAIT cutoff score. Overall, our findings confirmed our observations that a lower
161
CAIT cutoff score improved test characteristics, thus enhancing the usefulness of this patient
162
questionnaire in discriminating individuals with and without CAI
163
TE D
160
The CAIT is commonly used as either an inclusion criteria or descriptive tool for CAI subject populations.7-10 Because CAI as a pathology is classified symptomatically (as opposed to
165
using a diagnostic test such as an MRI as a “gold standard”), it is especially important to use
166
reliable and accurate methods in patient classification. The difference in cutoff scores between the current study and previous work can
AC C
167
EP
164
168
primarily be attributed to subject population definitions. As previously discussed, the original
169
calibration of the CAIT cutoff score used a history of ankle sprain alone to define group
170
membership, creating a group designation of “sprained” vs. “un-sprained” rather than a true
171
discrimination between “CAI” and “no CAI”. Sensations of giving way (a hallmark
172
characteristic of ankle instability4, 29) were not required in the original work by Hiller et al.14
8
ACCEPTED MANUSCRIPT
Thus, an individual with a history of sprain but no instability reporting a high CAIT score would
174
still have been categorized in the “sprained” group—elevating the average score in that group
175
and thus the optimal cutoff score. Additionally, a group of individuals in the original dataset
176
were dancers. It is possible that the skill level of these individuals may also have elevated
177
average CAIT scores in either or both groups.
178
Influence of “copers”
SC
179
RI PT
173
In addition to simply recalibrating and validating a new CAIT cutoff score, it was of particular interest in the current study to investigate how the test characteristics would be
181
affected by the inclusion of ankle sprain copers. These individuals in particular might be subject
182
to misclassification because they fall into the sprained group in a sprained vs. unsprained
183
paradigm, but into the no CAI group in a CAI vs. no CAI paradigm.
184
M AN U
180
Misclassification of a coper in the calibration dataset might falsely shift the cutoff score higher or lower. Thus we chose to calibrate the new cutoff score in a dataset which excluded
186
copers (Dataset 1), and validate the clinimetric properties of the new cutoff score in a dataset
187
which included copers (Dataset 2).
The ability of the CAIT to discriminate between CAI and uninjured controls in Dataset 1
EP
188
TE D
185
was very good. High sensitivity (95.1%) indicates that the CAIT would be an excellent screening
190
tool (few false negatives) to detect all possible cases. High specificity (100%) indicates that the
191
CAIT is also an excellent confirmation tool (no false positives).
192
AC C
189
LR+ and LR- are best applied clinically with a nomogram in situations where the pre-test
193
probability of disease is known. Using a conservative estimate from the work of Konradsen et
194
al.,3 the pre-test probability of an individual developing CAI post-ankle sprain is 32%.
195
Combining this data with the high LR+ (27.171) found in Dataset 1 leads to a post-test
9
ACCEPTED MANUSCRIPT
196
probability of approximately 90% (effectively ruling in CAI with a positive test), and the low
197
LR- (0.049) results to a post-test probability of approximately 2% (effectively ruling out CAI
198
with a negative test). We then desired to validate the recalibrated cutoff in an independent dataset which
200
included copers. Copers were included for two primary reasons. First, copers are now commonly
201
included as a comparison group in CAI research. Thus, it is of interest to investigate whether the
202
recalibrated CAIT cutoff would improve our ability to discriminate CAI from copers. Second,
203
even in research without a designated “coper” group it is possible that individuals with some
204
characteristics of copers could be inadvertently included if the CAI group definition was a
205
history of at least one ankle sprain and a CAIT score below the original cutoff (≤27). Inclusion of
206
copers within the CAI group would potentially alter or washout study results. Using our
207
validation Dataset 2, we found that a cutoff score of ≤23 yielded the highest Youden index
208
(0.893). However, the Youden index for the recalibrated cutoff score of ≤25 was only slightly
209
lower (0.834). The primary purpose of Dataset 2 was to validate if the specific recalibrated cutoff
210
score was appropriate in an independent dataset. With this intent in mind, the very small
211
difference in Youden index and clinimetric properties between the recalibrated cutoff from
212
Dataset 1 and the ideal cutoff in Dataset 2 provides evidence that the CAIT cutoff score for CAI
213
should indeed be at least ≤25. This finding agrees with the recommendation of the International
214
Ankle Consortium to lower the cutoff score used for inclusion into a CAI group,17 although it
215
disagrees on the exact recommended cutoff value. The ideal cutoff score (as identified by the
216
Youden index) was slightly lower in our Dataset 2 as compared to Dataset 1 primarily due to the
217
inclusion of copers whose CAIT scores varied widely (range, 23 to 30). This further emphasizes
218
the need to use a recalibrated cutoff score when working with a coper population.
AC C
EP
TE D
M AN U
SC
RI PT
199
10
ACCEPTED MANUSCRIPT
219
Caution should still be taken when using the recalibrated score with copers, as these individual had a high rate of false positives in the current study. Individuals working
221
intentionally with copers may be wise to elect a more conservative cutoff score (i.e. ≤23) for CAI
222
group membership, or elect to exclude copers whose CAIT score falls beneath the cutoff value.
223
Despite this caution, the recalibrated cutoff in the current study results in fewer false positives
224
than the original cutoff score, demonstrating the improved ability of the recalibrated CAIT cutoff
225
to appropriately discriminate between CAI and no CAI.
226
Comparison of clinimetric properties
SC
RI PT
220
Comparing clinimetric characteristics between our recalibration Dataset 1 and our
228
validation Dataset 2 (Table 5), all properties except LR- had decreased performance in Dataset 2.
229
This is because Dataset 2 included a more diverse subject pool. We felt including subjects across
230
a spectrum of ankle instability (CAI, copers and uninjured individual) was important to obtain
231
clinimetric characteristics that would be true to real life research and/or clinical practice.
TE D
232
M AN U
227
However, it should be noted that even in the validation Dataset 2 the recalibrated CAIT score still demonstrated very good clinimetric properties: high sensitivity, high specificity, high
234
LR+ and low LR-. All properties improved when compared to the original cutoff (Table 5),
235
further supporting the use of the recalibrated CAIT score. Clinically, using a standard nomogram
236
and assuming a pre-test probability of 32%,3 the LR+ in Dataset 2 (7.318) leads to a post-test
237
probability of approximately 68%, and the low LR- (0.039) leads to a post-test probability of
238
approximately 1.5% (effectively ruling out CAI with a negative test).
239
Study limitations and Recommendations for Future Research
240 241
AC C
EP
233
The CAIT instrument has been used to define group membership and/or describe subject characteristics in part because CAI lacks an objective “gold standard” test. Our reason to
11
ACCEPTED MANUSCRIPT
calculate a cutoff score is to add credence to the inclusion/exclusion of individuals into the CAI
243
or no CAI group. Ironically, using ROC curves to calculate a cutoff score requires that the
244
included subjects first be assigned group membership. We assigned that initial group
245
membership using what we felt were the most common and acceptable criteria at the time we
246
conducted the research.7, 13, 30 However, the values calculated in the current study are specific to
247
these definitions and any errors or philosophical disagreement with the original group
248
designation would affect clinimetric characteristics. For example, stricter criteria for coper group
249
inclusion might have led to copers with higher CAIT scores, which in turn may have resulted in a
250
higher ideal cutoff value in Dataset 2.
SC
M AN U
251
RI PT
242
Statistical comparison of subject weight between groups in Dataset 1 revealed that CAI subjects were significantly heavier than uninjured subjects. Data from three independent studies
253
were included in Dataset 1, thus the current study design does not facilitate an explanation of
254
why subjects with CAI were heavier. However, research by Hiller et al.31 on the prevalence and
255
impact of chronic musculoskeletal ankle disorders in the community found that 54.8% of
256
individuals with CAI reported limiting or modifying physical activity because of the ankle
257
problem. While this research did not report participant weight, nor a direct correlation between
258
CAI and weight, it might be expected that there is a link between limited physical activity and
259
increased body weight in this population. Future research on the health impact of CAI should
260
investigate this potential relationship more directly.
EP
AC C
261
TE D
252
Additionally, to validate our cutoff score we used an independent Datset 2 which did not
262
use CAIT score as an inclusion/exclusion factor. We felt this was important to obtain a truer
263
assessment of the clinimetric characteristics in real world situations. This led to inclusion of
264
copers who had a range of CAIT scores (23 to 30), which are wider than might be expected.
12
ACCEPTED MANUSCRIPT
Some individuals might disagree with assigning the label “coper” to an individual with a CAIT
266
score at the lower end of that range. Yet these individuals met our definition of a coper (a history
267
of a single lateral ankle sprain which required protected weight bearing, immobilization, and/or
268
limited activity for ≥ 24 hours, no subsequent re-sprains, had resumed all pre-injury activities
269
without limitation for at least 12 months prior to testing, and answer no when asked “Does your
270
ankle ever give-way, roll-over or feel unstable?”). This limitation further emphasizes the need to
271
include an established, reliable measure of patient reported symptoms within the subject
272
definition. Recent work by Donahue et al.13 also supports this need, as well as highlights the
273
ability of the CAIT (as compared to several other instability measures) to predict group
274
membership. Future research and clinical work involving copers should consider adding a score
275
of >25 on the CAIT as an inclusion criterion. Alternatively, future research could develop and
276
validate another reliable measure of patient reported symptoms that can be used for coper subject
277
classification.
278
Conclusion
SC
M AN U
TE D
279
RI PT
265
Clinicians and researchers using the CAIT to designate “CAI” vs. “no CAI” subject groups should use the recalibrated and validated CAIT cutoff score of ≤25 when assessing for
281
the presence or absence of CAI. This new cutoff score optimizes the clinimetric characteristics,
282
resulting in more accurate subject classification, and thereby assisting clinicians in their choice
283
of prevention and treatment strategies. Furthermore, the use of this new cut-off score in research
284
may result in a more accurate reflection of the CAI population. However, clinicians and
285
researchers alike should take caution when using this score with copers due to the high rate of
286
false positives in this population.
AC C
EP
280
13
ACCEPTED MANUSCRIPT
REFERENCES
288 289
1. Anandacoomarasamy A, Barnsley L. Long term outcomes of inversion ankle injuries. Br J Sports Med 2005;39(3):14-7.
290 291
2. Braun BL. Effects of ankle sprain in a general clinic population 6 to 18 months after medical evaluation. Arch Fam Med 1999;8(2):143-8.
292 293
3. Konradsen L, Bech L, Ehrenbjerg M, Nickelsen T. Seven years follow-up after ankle inversion trauma. Scand J Med Sci Sports 2002;12(3):129-35.
294 295
4. Hertel J. Functional anatomy, pathomechanics, and pathophysiology of lateral ankle instability. J Athl Train 2002;37(4):364-75.
296 297
5. Hiller CE, Kilbreath SL, Refshauge KM. Chronic ankle instability: Evolution of the model. Journal of Athletic Training 2011;46(2):133-41.
298 299
6. Delahunt E, Coughlan GF, Caulfield B, Nightingale EJ, Lin CW, Hiller CE. Inclusion criteria when investigating insufficiencies in chronic ankle instability. Med Sci Sports Exerc 2010;42(11):2106-21.
300 301 302
7. Wright CJ, Arnold BL, Ross SE, Ketchum JM, Ericksen JJ, Pidcoe PE. Clinical exam results differ among individuals with functional ankle instability and ankle sprain copers. J Athl Train 2012;In Press.
303 304
8. de Noronha M, Refshauge KM, Crosbie J, Kilbreath SL. Relationship between functional ankle instability and postural control. J Orthop Sports Phys Ther 2008;38(12):782-9.
305 306
9. Arnold BL, Wright CJ, Ross SE. Functional ankle instability and health-related quality of life. J Athl Train 2011;46(6):634-41.
307 308 309
10. Shields CA, Needle AR, Rose WC, Swanik CB, Kaminski TW. Effect of elastic taping on postural control deficits in subjects with healthy ankles, copers, and individuals with functional ankle instability. Foot Ankle Int 2013;34(10):1427-35.
310 311 312
11. De Noronha M, Refshauge KM, Kilbreath SL, Figueiredo VG. Cross-cultural adaptation of the brazilian-portuguese version of the cumberland ankle instability tool (CAIT). Disabil Rehabil 2008;30(26):1959-65.
313 314 315
12. Cruz-Díaz D, Hita-Contreras F, Lomas-Vega R, Osuna-Pérez MC, Martínez-Amat A. Cross-cultural adaptation and validation of the spanish version of the cumberland ankle instability tool (CAIT): An instrument to assess unilateral chronic ankle instability. Clin Rheumatol 2013;32(1):91-8.
316 317
13. Donahue M, Simon J, Docherty CL. Critical review of self-reported functional ankle instability measures. Foot Ankle Int 2011;32(12):1140-6.
318 319
14. Hiller CE, Refshauge KM, Bundy AC, Herbert RD, Kilbreath SL. The cumberland ankle instability tool: A report of validity and reliability testing. Arch Phys Med Rehabil 2006;87(9):1235-41.
320 321
15. Hiller CE, Refshauge KM, Herbert RD, Kilbreath SL. Balance and recovery from a perturbation are impaired in people with functional ankle instability. Clin J Sport Med 2007;17(4):269-75.
AC C
EP
TE D
M AN U
SC
RI PT
287
14
ACCEPTED MANUSCRIPT
16. de Noronha M, Refshauge KM, Kilbreath SL, Crosbie J. Loss of proprioception or motor control is not related to functional ankle instability: An observational study. Aust J Physiother 2007;53(3):193-8.
324 325 326 327
17. Gribble PA, Delahunt E, Bleakley C, Caulfield B, Docherty CL, Fourchet F, Fong D, Hertel J, Hiller C, Kaminski TW, et al. Selection criteria for patients with chronic ankle instability in controlled research: A position statement of the international ankle consortium. J Orthop Sports PhyS Ther 2013;43(8):585-91.
328 329 330
18. Brown C, Padua D, Marshall SW, Guskiewicz K. Individuals with mechanical ankle instability exhibit different motion patterns than those with functional ankle instability and ankle sprain copers. Clin Biomech 2008;23(6):822-31.
331 332
19. Hubbard TJ. Ligament laxity following inversion injury with and without chronic ankle instability. Foot Ankle Int 2008;29(3):305-11.
333 334 335
20. Wikstrom EA, Tillman MD, Chmielewski TL, Cauraugh JH, Naugle KE, Borsa PA. Self-assessed disability and functional performance in individuals with and without ankle instability: A case control study. J Orthop Sports Phys Ther 2009;39(6):458-67.
336 337
21. Wikstrom EA, Fournier KA, McKeon PO. Postural control differs between those with and without chronic ankle instability. Gait Posture 2010;32(1):82-6.
338 339 340
22. Wikstrom EA, Tillman MD, Chmielewski TL, Cauraugh JH, Naugle KE, Borsa PA. Dynamic postural control but not mechanical stability differs among those with and without chronic ankle instability. Scand J Med Sci Sports 2010;20(1):e137-44.
341 342
23. Brown C. Foot clearance in walking and running in individuals with ankle instability. Am J Sports Med 2011;39(8):1769-76.
343 344
24. Wikstrom, E.A., Hass, C.J. Gait termination strategies differ between those with and without ankle instability. Clin Biomech 2012;27(6):619-24.
345 346 347
25. Croy T, Saliba SA, Saliba E, Anderson MW, Hertel J. Differences in lateral ankle laxity measured via stress ultrasonography in individuals with chronic ankle instability, ankle sprain copers, and healthy individuals. J Orthop Sports Phys Ther 2012;42(7):593-600.
348 349
26. Wikstrom EA, Tillman MD, Chmielewski TL, Cauraugh JH, Naugle KE, Borsa PA. Discriminating between copers and people with chronic ankle instability. J Athl Train 2012;47(2):136-42.
350
27. Youden WJ. Index for rating diagnostic tests. Cancer 1950;3(1):32-5.
351 352 353
28. Jaeschke R, Guyatt GH, Sackett DL. Users' guides to the medical literature. III. how to use an article about a diagnostic test. B. what are the results and will they help me in caring for my patients? the evidence-based medicine working group. JAMA 1994;271(9):703-7.
354 355
29. Freeman MR, Dean ME, Hanham IF. The etiology and prevention of functional instability of the foot. J Bone Joint Surg Br 1965;47(4):678-85.
356 357
30. Arnold BL, Linens SW, de la Motte SJ, Ross SE. Concentric evertor strength differences are associated with functional ankle instability: A meta-analysis. J Athl Train 2009;44(6):653-62.
AC C
EP
TE D
M AN U
SC
RI PT
322 323
15
ACCEPTED MANUSCRIPT
EP
TE D
M AN U
SC
RI PT
31. Hiller CE, Nightingale EJ, Raymond J, Kilbreath SL, Burns J, Black DA, Refshauge KM. Prevalence and impact of chronic musculoskeletal ankle disorders in the community. Arch Phys Med Rehabil 2012;93(10):1801-7.
AC C
358 359 360
16
ACCEPTED MANUSCRIPT
361
LEGEND TO FIGURES
362 Figure 1. Receiver operating characteristic (ROC) curve for Dataset 1 Cumberland Ankle Instability Tool
364
(CAIT) scores. Solid line = ROC curve, Doted line = reference line for significant ROC curve. *Cutoff value
365
with highest Youden Index
RI PT
363
366
Figure 2. Receiver operating characteristic (ROC) curve for Dataset 2 Cumberland Ankle Instability Tool
368
(CAIT) scores. Solid line = ROC curve, Doted line = reference line for significant ROC curve. *Cutoff value
369
with highest Youden Index in Dataset 2. Cutoff value identified in Dataset 1.
SC
367
†
AC C
EP
TE D
M AN U
370
17
ACCEPTED MANUSCRIPT
371
LEGEND TO TABLES
372 373
Table 1. Subject Demographics Dataset 1
375
Table 2. Subject Demographics Dataset 2
376 377
Table 3. Clinimetric properties at each potential CAIT cutoff using Dataset 1
SC
378
RI PT
374
Table 4. Clinimetric properties at each potential CAIT cutoff using Dataset 2
380
Table 5. Comparison of CAIT clinimetric properties between datasets
AC C
EP
TE D
M AN U
379
18
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Table 1. Subject Demographics Dataset 1 Descriptor Uninjured t-statistic df P-value CAI Gender 26 males 20 males, 35 females 37 females Age, yrs 25.52 ± 6.31 25.02 ± 5.49 -0.464 116 0.643 Height, m 1.71 ± 0.08 1.69 ± 0.08 -1.057 116 0.293 Weight, kg 77.07 ± 16.02 69.16 ± 13.06 -2.931 116 0.004* Abbreviations: CAI = Chronic Ankle Instability, CAIT = Cumberland Ankle Instability Tool. Values are mean ± standard deviation. *Significant difference in weight between CAI and uninjured subjects
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Table 2. Subject Demographics Dataset 2 Descriptor Coper Uninjured F-statistic P-Value CAI Gender 15 males, 12 males, 14 males, 14 females 14 females 13 females Age, yrs 23.31 ± 3.53 23.35 ± 3.50 22.89 ± 3.78 F2,79=0.135 0.874 0.847 Height, m 1.72 ± 0.10 1.71 ± 0.07 1.71 ± 0.08 F2,79=0.166 Weight, kg 75.12 ± 19.52 69.81 ± 13.65 69.33 ± 13.90 F2,79=1.128 0.329 Abbreviations: CAI = Chronic Ankle Instability, CAIT = Cumberland Ankle Instability Tool. Values are mean ± standard deviation.
ACCEPTED MANUSCRIPT
RI PT
LR1.000 0.984 0.967 0.951 0.934 0.902 0.836 0.754 0.672 0.623 0.525 0.410 0.344 0.246 0.180 0.082 0.049 0.051 0.017 0.000 0.000 -
TE D
M AN U
SC
Table 3. Clinimetric properties at each potential CAIT cutoff using Dataset 1. CAIT Sn 1-Sp Sp Youden Index LR+ 7.0 0.000 0.000 1.000 0.000 8.5 0.016 0.000 1.000 0.016 9.5 0.033 0.000 1.000 0.033 11.0 0.049 0.000 1.000 0.049 13.0 0.066 0.000 1.000 0.066 14.5 0.098 0.000 1.000 0.098 15.5 0.164 0.000 1.000 0.164 16.5 0.246 0.000 1.000 0.246 17.5 0.328 0.000 1.000 0.328 18.5 0.377 0.000 1.000 0.377 19.5 0.475 0.000 1.000 0.475 20.5 0.590 0.000 1.000 0.590 21.5 0.656 0.000 1.000 0.656 22.5 0.754 0.000 1.000 0.754 23.5 0.820 0.000 1.000 0.820 24.5 0.918 0.000 1.000 0.918 25.5 0.951 0.000 1.000 0.951 26.5 0.951 0.035 0.965 0.916 27.171 27.5 0.984 0.070 0.930 0.914 14.057 28.5 1.000 0.158 0.842 0.842 6.329 29.5 1.000 0.211 0.789 0.789 4.739 31.0 1.000 1.000 0.000 0.000 1.000
AC C
EP
Abbreviations: CAIT = Cumberland Ankle Instability Tool, Sn = sensitivity, Sp = Specificity, LR+ = Positive likelihood ratio, LR- = Negative Likelihood ratio
ACCEPTED MANUSCRIPT
RI PT
LR1.000 0.966 0.931 0.897 0.793 0.759 0.690 0.621 0.517 0.345 0.241 0.072 0.073 0.039 0.000 0.000 0.000 0.000 -
M AN U
SC
Table 4. Clinimetric properties at each potential CAIT cutoff using Dataset 2. CAIT Sn 1-Sp Sp Youden Index LR+ 5.0 0.000 0.000 1.000 0.000 9.0 0.034 0.000 1.000 0.034 13.0 0.069 0.000 1.000 0.069 15.0 0.103 0.000 1.000 0.103 16.5 0.207 0.000 1.000 0.207 17.5 0.241 0.000 1.000 0.241 18.5 0.310 0.000 1.000 0.310 19.5 0.379 0.000 1.000 0.379 20.5 0.483 0.000 1.000 0.483 21.5 0.655 0.000 1.000 0.655 22.5 0.759 0.000 1.000 0.759 23.5 0.931 0.038 0.962 0.893 24.500 24.5 0.931 0.057 0.943 0.874 16.333 25.5 0.966 0.132 0.868 0.834 7.318 26.5 1.000 0.189 0.811 0.811 5.291 27.5 1.000 0.458 0.542 0.542 2.183 28.5 1.000 0.472 0.528 0.528 2.119 29.5 1.000 0.623 0.377 0.377 1.605 31.00 1.000 1.000 0.000 0.000 1.000
AC C
EP
TE D
Abbreviations: CAIT = Cumberland Ankle Instability Tool, Sn = sensitivity, Sp = Specificity, LR+ = Positive likelihood ratio, LR- = Negative Likelihood ratio
ACCEPTED MANUSCRIPT
Table 5. Comparison of CAIT clinimetric properties between datasets Hiller 2006 Dataset 1 Test Characteristic (cutoff: ≤27) (cutoff: ≤25)
Dataset 2 (cutoff: ≤25)
AC C
EP
TE D
M AN U
SC
RI PT
Sensitivity 0.860 0.951 0.966 Specificity 0.830 1.000 0.868 a Positive Likelihood ratio 4.890 27.171 7.318 Negative Likelihood ratio 0.180 0.051 0.039 Abbreviations: CAIT = Cumberland Ankle Instability Tool a Due to perfect specificity, the positive likelihood ratio could not be calculated at the ≤25 cutoff, the value given is for the next closest value (≤26) where it could be calculated.
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
25.5*
AC C
EP
TE D
Figure 1.
ACCEPTED MANUSCRIPT
25.5†
M AN U
SC
RI PT
23.5*
AC C
EP
TE D
Figure 2.
S0003-9993(14)00334-7
DOI:
10.1016/j.apmr.2014.04.017
Reference:
YAPMR 55822
To appear in:
ARCHIVES OF PHYSICAL MEDICINE AND REHABILITATION
Received Date: 26 November 2013 Revised Date:
14 April 2014
Accepted Date: 17 April 2014
Please cite this article as: Wright CJ, Arnold BL, Ross SE, Linens SW, Recalibration and validation of the Cumberland Ankle Instability Tool cutoff score for individuals with Chronic Ankle Instability, ARCHIVES OF PHYSICAL MEDICINE AND REHABILITATION (2014), doi: 10.1016/ j.apmr.2014.04.017. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Running Head: Recalibration of the CAIT in CAI individuals
Authors name and highest academic degree (in order of authorship) Cynthia J Wright*, PhD, ATC Brent L Arnold†, PhD, ATC Scott E Ross‡, PhD, ATC Shelley W Linens§, PhD, ATC
RI PT
Title: Recalibration and validation of the Cumberland Ankle Instability Tool cutoff score for individuals with Chronic Ankle Instability
M AN U
SC
Research was conducted in the Department of Health and Human Performance at Virginia Commonwealth University, Richmond, Virginia, USA. All authors were affiliated with Virginia Commonwealth University at the time of study completion. All authors have changed affiliations since study completion. Current affiliations are indicated below. * Whitworth University † Indiana University School of Health and Rehabilitation Sciences ‡ University of North Caroline Greensboro § Georgia State University
TE D
Portions of the content of this manuscript were presented at 2 national conferences: Wright CJ, Arnold BL, Ross SE, Pidcoe PE. (2011, June). Validation of a Recalibrated Cumberland Ankle Instability Tool Cutoff Score for Chronic Ankle Instability. Poster session for the 2011 National Athletic Trainer’s Association Annual Meeting and Clinical Symposium, New Orleans, LA. Published Abstract: J Athl Train 2011: 46(3)(Suppl):S124. Arnold BL, Wright CJ, Linens SW, Ross SE (2010, June). Recalibration of the CAIT cutoff score for chronic ankle instability. Poster presentation for the 2011 American College of Sports Medicine Annual Meeting, Denver, CO. Published Abstract: Med Sci Sports Exerc 2011: 43(5)(Suppl)S341.
EP
No financial support was received for this research. The authors have no conflicts of interest to disclose.
AC C
Shelley Linens reports a doctoral research grant from the National Athletic Trainers' Association Research and Education Foundation Corresponding author: Cynthia Wright Whitworth University 300 W Hawthorne Road Spokane, WA 99251 (509) 777-3244 Office (509) 777-4943 Fax [email protected]
ACCEPTED MANUSCRIPT
1
Recalibration and validation of the Cumberland Ankle Instability Tool cutoff score
2
for individuals with Chronic Ankle Instability
3
Objective: To independently recalibrate and re-validate the Cumberland Ankle Instability Tool
5
(CAIT) cutoff score for discriminating individuals with and without chronic ankle instability
6
(CAI). There are concerns the original cutoff score (≤27) may be suboptimal for use in the CAI
7
population. Design: Case-control. Setting: Research laboratory. Participants: Two independent
8
datasets were used. Dataset 1 included 61 individuals with a history of ≥1 ankle sprain and ≥2
9
episodes of giving-way in the past year (CAI), and 57 participants with no history of ankle sprain
M AN U
SC
RI PT
4
or instability in their lifetime (uninjured). Dataset 2 included 27 uninjured participants, 29 CAI
11
participants, and 26 individuals with a history of a single ankle sprain and no subsequent
12
instability (copers). Interventions: All participants completed the CAIT during a single session.
13
In Dataset 1 a receiver operating curve (ROC) was calculated using CAIT score and group
14
membership as the test variables. The ideal cutoff score was identified using Youden’s index.
15
The recalibrated cutoff score was validated in Dataset 2 using ROC analysis and clinimetric
16
characteristics. Main Outcome Measure(s): CAIT cutoff score and clinimetrics. Results: In
17
Dataset 1, the optimal cutoff score was ≤25, which is lower than previously reported. In Dataset
18
2 the recalibrated cutoff score demonstrated a sensitivity of 96.6%, specificity of 86.8%, positive
19
likelihood ratio (LR+) of 7.318, negative likelihood ratio (LR-) of 0.039. There were seven false
20
positives and one false negative. Conclusions: The recalibrated CAIT score demonstrated very
21
good clinimetric properties; all properties improved when compared to the original cutoff.
22
Clinicians using the CAIT should utilize the recalibrated cutoff score to maximize test
23
characteristics. Caution should be taken with copers, who had a high rate of false positives.
AC C
EP
TE D
10
1
ACCEPTED MANUSCRIPT
24 25
Keywords: Functional Ankle Instability; Ankle Sprain; Clinimetrics; Patient Questionnaire;
26
Self-Reported Measure; Test Characteristics
28
List of abbreviations:
29
CAI
30
CAIT Cumberland Ankle Instability Tool
31
ROC receiver operating curve
32
LR+
positive likelihood ratio
33
LR-
negative likelihood ratio
34
AUC area under the curve
RI PT
27
AC C
EP
TE D
M AN U
SC
Chronic Ankle Instability
2
ACCEPTED MANUSCRIPT
35
Chronic ankle instability (CAI) is a common sequelae of lateral ankle sprain, affecting approximately 32-47% of ankle sprain patients.1-3 A symptomatically defined pathology, CAI is
37
characterized by recurrent sprains and/or recurrent instability (e.g. episodes of “giving-way”)
38
after an ankle sprain.4, 5 Recent articles have reviewed the problems and variability involved with
39
a symptomatic definition of CAI.5, 6
40
RI PT
36
Patient questionnaires can serve various functions. One function can be to provide
reliable measurement of patient reported symptoms such as pain, functional limitations, and
42
instability occurrence with specific activities. A questionnaire that has been widely used in ankle
43
instability literature,7-10 translated into multiple languages,11, 12 and shown to be a significant
44
predictor of ankle instability status,13 is the Cumberland Ankle Instability Tool (CAIT).14 First
45
published by Hiller et al.14 in 2006, this nine question survey focuses on symptoms of instability
46
during several different physical tasks. The CAIT results in a score ranging from zero to 30 with
47
higher scores indicating higher stability. Original research established a cutoff score of ≤27 as
48
indicative of CAI group membership.14 We observed in our laboratory that the established cutoff
49
score appeared to be too high. Individuals who had a history of ankle sprain, but subjectively
50
reported that their ankle “didn’t really bother them” were occasionally classified as having CAI
51
based on the established cutoff score of ≤27. Perhaps because of this issue, some authors have
52
independently chosen to use a lower cutoff score (i.e. ≤23 and ≤24) than what was originally
53
validated by Hiller et al.15, 16 Recently, the International Ankle Consortium recommended that a
54
cutoff score of ≤24 be used in CAI inclusion criteria.17 However, to our knowledge statistical
55
evidence to support the selection of a lower cutoff value has not been reported in the literature.
56
Further investigation into the question of appropriate cutoff scores for the CAIT
57
AC C
EP
TE D
M AN U
SC
41
highlighted an important limitation in the criteria used to establish the original cutoff score.
3
ACCEPTED MANUSCRIPT
Hiller et al.14 used a history of ankle sprain alone to define group membership when calculating
59
the cutoff score. This created a group designation of “sprained” vs. “un-sprained”—yet the cutoff
60
score derived has been commonly used to define “CAI” vs. “No CAI” group membership. These
61
two categorization schemes have important differences, and we propose should not be used
62
interchangeably. Specifically, recent research has highlighted that some individuals (frequently
63
called “copers”) have a history of ankle sprain but no ongoing instability.7, 18-26 Copers score
64
similar to uninjured controls on questionnaires such as the Foot and Ankle Ability Measure
65
(FAAM), FAAM-Sport, and CAIT.7 Thus, inclusion of these individuals in the sprained group
66
could inflate scores and lead to the establishment of a higher cutoff value that would be
67
established if only symptomatic ankles were included.
SC
M AN U
68
RI PT
58
The purpose of the current investigation was to independently re-validate and potentially recalibrate a CAIT cutoff score, by including only individuals with a history of lateral ankle
70
sprain AND recurrent instability in the CAI group.13 We hypothesized that the resulting cutoff
71
score would be lower than previously reported. Additionally, we desired to test whether a
72
recalibrated cutoff score would appropriately classify subjects in an independent subject pool
73
including uninjured control subjects, copers and CAI individuals. We hypothesized that a lower
74
cutoff score would result in fewer false classifications of copers.
EP
AC C
75
TE D
69
76
METHODS
77
Subjects Dataset 1: Recalibration
78
Subjects were originally recruited for three independent research studies which collected
79
CAIT scores. Participants were recruited from a large metropolitan area, including a university
80
campus. The one hundred and eighteen individuals were recruited via direct contact with
4
ACCEPTED MANUSCRIPT
individuals and recruiting announcements in university courses and included in this study: 61
82
individuals with CAI and 57 uninjured individuals. Demographics are reported in Table 1.
83
Inclusion criteria for the CAI group included a history of at least one lateral ankle sprain >6
84
weeks prior to study entry and at least two reported episodes of giving way per year. Uninjured
85
individuals had no history of ankle sprain or instability in their lifetime. Exclusion criteria for
86
both groups included a history of lower extremity fracture or surgery, any acute symptoms of
87
ankle injury, or assisted ambulation. University Institutional Review Board approval was
88
obtained prior to data collection for both datasets.
89
Subjects Dataset 2: Validation
SC
M AN U
90
RI PT
81
Eighty-two subjects were recruited from a large metropolitan area, including a university campus via direct contact with individuals and recruiting announcements made in university
92
courses. Twenty-nine individuals with CAI, 26 copers and 27 uninjured individuals were
93
included in the dataset. Table 2 includes subject demographics. Inclusion criteria for the CAI
94
group and uninjured group were the same as in Dataset 1, with the addition that CAI subjects had
95
to be at least 1 year post-initial injury. Individuals categorized as copers reported a history of a
96
single lateral ankle sprain which required protected weight bearing, immobilization, and/or
97
limited activity for ≥ 24 hours, no perceived instability, and had resumed all pre-injury activities
98
without limitation for at least 12 months prior to testing. Perceived instability was assessed with
99
a single yes or no question, “Does your ankle ever give-way, roll-over or feel unstable?”
AC C
EP
TE D
91
100
Exclusion criteria were the same as in Dataset1. Additionally, subjects had to perform at least 90
101
minutes of physical activity per week; this activity could be of any intensity or mode. The CAIT
102
score was not used as an inclusion criteria for any group.
103
Data Collection
5
ACCEPTED MANUSCRIPT
104
In each study, subjects reported to the University Sports Medicine Research Laboratory and gave informed consent. Inclusion and exclusion criteria were verified, demographic data
106
(including age, height and weight) was collected, and the CAIT was completed. A custom
107
computer program (Access, Microsoft, Redmond, WA) recorded and scored CAIT questionnaire
108
responses for each subject. Only CAIT scores for the involved limb (CAI subjects) or
109
comparison limb (uninjured subjects) were utilized in the analysis. For CAI individuals with
110
bilateral instability, the most unstable ankle (i.e. lowest CAIT score) was included in the
111
analysis.
112
Statistical Analysis
SC
M AN U
113
RI PT
105
All analyses were performed using IBM SPSS Statistics 20 (Armonk, New York, USA). Paired t-tests were used to compare subject demographics between groups in Dataset 1, and a
115
one-way ANOVA was used to compare the same variables among groups in Dataset 2. For both
116
datasets, receiver operator characteristic (ROC) curves were generated with CAIT score as the
117
dependent variable, and group membership (CAI vs. no CAI) as the independent variable. In
118
Dataset 2, the no CAI group included both copers and uninjured subjects. Area under the curve
119
(AUC) was used to identify a significant ROC curve using a one-sided test (alpha = 0.05).
120
After calculating the ROC curve, the diagnostic sensitivity and specificity for each
EP
TE D
114
potential cutoff score were calculated. The largest Youden index value [sensitivity + specificity-
122
1)] was used to determine the ideal cutoff score.27 In Dataset 1, the clinical meaningfulness of
123
each cutoff score was evaluated by calculating the positive likelihood ratio (LR+) and negative
124
likelihood ratio (LR-). LR+ was calculated as [sensitivity/(1-specficity)], and LR- was calculated
125
as [(1-sensitivity)/specificity]. In Dataset 2, the purpose was to validate the recalibrated cutoff
126
score (≤25), thus clinimetric properties including the diagnostic sensitivity and specificity were
AC C
121
6
ACCEPTED MANUSCRIPT
calculated for the recalibrated cutoff score. Additionally, the clinical meaningfulness of the
128
recalibrated cutoff score was evaluated by calculating the LR+, LR-, number of false positives
129
and number of false negatives. A low LR- value (<0.2) would indicate that a negative test
130
substantially decreases the likelihood of an individual truly having CAI, a high LR+ (>5) would
131
indicate that a positive test substantially increases the likelihood of an individual truly having
132
CAI.28
SC
RI PT
127
134
RESULTS
135
Dataset 1: Recalibration
136
M AN U
133
The average CAIT score for the uninjured subject group was 29.53±1.04 (range, 26 to 30), and the CAI group was 19.41±4.27 (range, 8 to 28). Results of statistical tests on
138
demographic variables are reported in Table 1. The ROC was significant (AUC = 0.996,
139
p=0.005; Figure 1), indicating that CAIT score significantly predicted group membership. The
140
largest Youden index value (0.95) indicated that a CAIT score ≤25 was the ideal cutoff to
141
distinguish group membership (Table 3). High sensitivity (95.1%) and specificity (100%) were
142
calculated at this cutoff (Table 3). The LR- value was 0.049. Due to perfect specificity, a LR+
143
could not be calculated at the recalibrated cutoff score. However, the next nearest cutoff value
144
where LR+ could be calculated (26.5) resulted in a LR+ of 27.171.
145
Dataset 2: Validation
EP
AC C
146
TE D
137
The average CAIT scores by group were 28.93±1.69 (range, 23 to 30) for the uninjured
147
subjects, 27.31±2.02 (range, 23 to 30) for copers, and 19.59±4.15 (range, 6 to 26) for CAI
148
subjects. Results of statistical tests on demographic variables are reported in Table 2. The ROC
149
was significant (AUC = 0.988, p<0.001; Figure 2), indicating that CAIT score again significantly
7
ACCEPTED MANUSCRIPT
150
predicted group membership. The largest Youden index value (0.893) indicated that a CAIT
151
score of ≤23 was the ideal cutoff to distinguish group membership in this dataset (Table 4). The
152
Youden index value for the recalibrated cutoff score of ≤25 was only slightly lower (0.834). At the recalibrated cutoff score of ≤25, sensitivity (96.6%) and specificity (86.6%) were
RI PT
153
both high (Table 4 and Table 5). The LR- value was 0.039, and the LR+ was 7.318. Using the
155
recalibrated cutoff score there were seven false positives (one uninjured subject and six copers)
156
and one false negative (one CAI subject).
SC
154
157
159
DISCUSSION
M AN U
158
The purpose of the current investigation was to independently re-validate and potentially recalibrate a CAIT cutoff score. Overall, our findings confirmed our observations that a lower
161
CAIT cutoff score improved test characteristics, thus enhancing the usefulness of this patient
162
questionnaire in discriminating individuals with and without CAI
163
TE D
160
The CAIT is commonly used as either an inclusion criteria or descriptive tool for CAI subject populations.7-10 Because CAI as a pathology is classified symptomatically (as opposed to
165
using a diagnostic test such as an MRI as a “gold standard”), it is especially important to use
166
reliable and accurate methods in patient classification. The difference in cutoff scores between the current study and previous work can
AC C
167
EP
164
168
primarily be attributed to subject population definitions. As previously discussed, the original
169
calibration of the CAIT cutoff score used a history of ankle sprain alone to define group
170
membership, creating a group designation of “sprained” vs. “un-sprained” rather than a true
171
discrimination between “CAI” and “no CAI”. Sensations of giving way (a hallmark
172
characteristic of ankle instability4, 29) were not required in the original work by Hiller et al.14
8
ACCEPTED MANUSCRIPT
Thus, an individual with a history of sprain but no instability reporting a high CAIT score would
174
still have been categorized in the “sprained” group—elevating the average score in that group
175
and thus the optimal cutoff score. Additionally, a group of individuals in the original dataset
176
were dancers. It is possible that the skill level of these individuals may also have elevated
177
average CAIT scores in either or both groups.
178
Influence of “copers”
SC
179
RI PT
173
In addition to simply recalibrating and validating a new CAIT cutoff score, it was of particular interest in the current study to investigate how the test characteristics would be
181
affected by the inclusion of ankle sprain copers. These individuals in particular might be subject
182
to misclassification because they fall into the sprained group in a sprained vs. unsprained
183
paradigm, but into the no CAI group in a CAI vs. no CAI paradigm.
184
M AN U
180
Misclassification of a coper in the calibration dataset might falsely shift the cutoff score higher or lower. Thus we chose to calibrate the new cutoff score in a dataset which excluded
186
copers (Dataset 1), and validate the clinimetric properties of the new cutoff score in a dataset
187
which included copers (Dataset 2).
The ability of the CAIT to discriminate between CAI and uninjured controls in Dataset 1
EP
188
TE D
185
was very good. High sensitivity (95.1%) indicates that the CAIT would be an excellent screening
190
tool (few false negatives) to detect all possible cases. High specificity (100%) indicates that the
191
CAIT is also an excellent confirmation tool (no false positives).
192
AC C
189
LR+ and LR- are best applied clinically with a nomogram in situations where the pre-test
193
probability of disease is known. Using a conservative estimate from the work of Konradsen et
194
al.,3 the pre-test probability of an individual developing CAI post-ankle sprain is 32%.
195
Combining this data with the high LR+ (27.171) found in Dataset 1 leads to a post-test
9
ACCEPTED MANUSCRIPT
196
probability of approximately 90% (effectively ruling in CAI with a positive test), and the low
197
LR- (0.049) results to a post-test probability of approximately 2% (effectively ruling out CAI
198
with a negative test). We then desired to validate the recalibrated cutoff in an independent dataset which
200
included copers. Copers were included for two primary reasons. First, copers are now commonly
201
included as a comparison group in CAI research. Thus, it is of interest to investigate whether the
202
recalibrated CAIT cutoff would improve our ability to discriminate CAI from copers. Second,
203
even in research without a designated “coper” group it is possible that individuals with some
204
characteristics of copers could be inadvertently included if the CAI group definition was a
205
history of at least one ankle sprain and a CAIT score below the original cutoff (≤27). Inclusion of
206
copers within the CAI group would potentially alter or washout study results. Using our
207
validation Dataset 2, we found that a cutoff score of ≤23 yielded the highest Youden index
208
(0.893). However, the Youden index for the recalibrated cutoff score of ≤25 was only slightly
209
lower (0.834). The primary purpose of Dataset 2 was to validate if the specific recalibrated cutoff
210
score was appropriate in an independent dataset. With this intent in mind, the very small
211
difference in Youden index and clinimetric properties between the recalibrated cutoff from
212
Dataset 1 and the ideal cutoff in Dataset 2 provides evidence that the CAIT cutoff score for CAI
213
should indeed be at least ≤25. This finding agrees with the recommendation of the International
214
Ankle Consortium to lower the cutoff score used for inclusion into a CAI group,17 although it
215
disagrees on the exact recommended cutoff value. The ideal cutoff score (as identified by the
216
Youden index) was slightly lower in our Dataset 2 as compared to Dataset 1 primarily due to the
217
inclusion of copers whose CAIT scores varied widely (range, 23 to 30). This further emphasizes
218
the need to use a recalibrated cutoff score when working with a coper population.
AC C
EP
TE D
M AN U
SC
RI PT
199
10
ACCEPTED MANUSCRIPT
219
Caution should still be taken when using the recalibrated score with copers, as these individual had a high rate of false positives in the current study. Individuals working
221
intentionally with copers may be wise to elect a more conservative cutoff score (i.e. ≤23) for CAI
222
group membership, or elect to exclude copers whose CAIT score falls beneath the cutoff value.
223
Despite this caution, the recalibrated cutoff in the current study results in fewer false positives
224
than the original cutoff score, demonstrating the improved ability of the recalibrated CAIT cutoff
225
to appropriately discriminate between CAI and no CAI.
226
Comparison of clinimetric properties
SC
RI PT
220
Comparing clinimetric characteristics between our recalibration Dataset 1 and our
228
validation Dataset 2 (Table 5), all properties except LR- had decreased performance in Dataset 2.
229
This is because Dataset 2 included a more diverse subject pool. We felt including subjects across
230
a spectrum of ankle instability (CAI, copers and uninjured individual) was important to obtain
231
clinimetric characteristics that would be true to real life research and/or clinical practice.
TE D
232
M AN U
227
However, it should be noted that even in the validation Dataset 2 the recalibrated CAIT score still demonstrated very good clinimetric properties: high sensitivity, high specificity, high
234
LR+ and low LR-. All properties improved when compared to the original cutoff (Table 5),
235
further supporting the use of the recalibrated CAIT score. Clinically, using a standard nomogram
236
and assuming a pre-test probability of 32%,3 the LR+ in Dataset 2 (7.318) leads to a post-test
237
probability of approximately 68%, and the low LR- (0.039) leads to a post-test probability of
238
approximately 1.5% (effectively ruling out CAI with a negative test).
239
Study limitations and Recommendations for Future Research
240 241
AC C
EP
233
The CAIT instrument has been used to define group membership and/or describe subject characteristics in part because CAI lacks an objective “gold standard” test. Our reason to
11
ACCEPTED MANUSCRIPT
calculate a cutoff score is to add credence to the inclusion/exclusion of individuals into the CAI
243
or no CAI group. Ironically, using ROC curves to calculate a cutoff score requires that the
244
included subjects first be assigned group membership. We assigned that initial group
245
membership using what we felt were the most common and acceptable criteria at the time we
246
conducted the research.7, 13, 30 However, the values calculated in the current study are specific to
247
these definitions and any errors or philosophical disagreement with the original group
248
designation would affect clinimetric characteristics. For example, stricter criteria for coper group
249
inclusion might have led to copers with higher CAIT scores, which in turn may have resulted in a
250
higher ideal cutoff value in Dataset 2.
SC
M AN U
251
RI PT
242
Statistical comparison of subject weight between groups in Dataset 1 revealed that CAI subjects were significantly heavier than uninjured subjects. Data from three independent studies
253
were included in Dataset 1, thus the current study design does not facilitate an explanation of
254
why subjects with CAI were heavier. However, research by Hiller et al.31 on the prevalence and
255
impact of chronic musculoskeletal ankle disorders in the community found that 54.8% of
256
individuals with CAI reported limiting or modifying physical activity because of the ankle
257
problem. While this research did not report participant weight, nor a direct correlation between
258
CAI and weight, it might be expected that there is a link between limited physical activity and
259
increased body weight in this population. Future research on the health impact of CAI should
260
investigate this potential relationship more directly.
EP
AC C
261
TE D
252
Additionally, to validate our cutoff score we used an independent Datset 2 which did not
262
use CAIT score as an inclusion/exclusion factor. We felt this was important to obtain a truer
263
assessment of the clinimetric characteristics in real world situations. This led to inclusion of
264
copers who had a range of CAIT scores (23 to 30), which are wider than might be expected.
12
ACCEPTED MANUSCRIPT
Some individuals might disagree with assigning the label “coper” to an individual with a CAIT
266
score at the lower end of that range. Yet these individuals met our definition of a coper (a history
267
of a single lateral ankle sprain which required protected weight bearing, immobilization, and/or
268
limited activity for ≥ 24 hours, no subsequent re-sprains, had resumed all pre-injury activities
269
without limitation for at least 12 months prior to testing, and answer no when asked “Does your
270
ankle ever give-way, roll-over or feel unstable?”). This limitation further emphasizes the need to
271
include an established, reliable measure of patient reported symptoms within the subject
272
definition. Recent work by Donahue et al.13 also supports this need, as well as highlights the
273
ability of the CAIT (as compared to several other instability measures) to predict group
274
membership. Future research and clinical work involving copers should consider adding a score
275
of >25 on the CAIT as an inclusion criterion. Alternatively, future research could develop and
276
validate another reliable measure of patient reported symptoms that can be used for coper subject
277
classification.
278
Conclusion
SC
M AN U
TE D
279
RI PT
265
Clinicians and researchers using the CAIT to designate “CAI” vs. “no CAI” subject groups should use the recalibrated and validated CAIT cutoff score of ≤25 when assessing for
281
the presence or absence of CAI. This new cutoff score optimizes the clinimetric characteristics,
282
resulting in more accurate subject classification, and thereby assisting clinicians in their choice
283
of prevention and treatment strategies. Furthermore, the use of this new cut-off score in research
284
may result in a more accurate reflection of the CAI population. However, clinicians and
285
researchers alike should take caution when using this score with copers due to the high rate of
286
false positives in this population.
AC C
EP
280
13
ACCEPTED MANUSCRIPT
REFERENCES
288 289
1. Anandacoomarasamy A, Barnsley L. Long term outcomes of inversion ankle injuries. Br J Sports Med 2005;39(3):14-7.
290 291
2. Braun BL. Effects of ankle sprain in a general clinic population 6 to 18 months after medical evaluation. Arch Fam Med 1999;8(2):143-8.
292 293
3. Konradsen L, Bech L, Ehrenbjerg M, Nickelsen T. Seven years follow-up after ankle inversion trauma. Scand J Med Sci Sports 2002;12(3):129-35.
294 295
4. Hertel J. Functional anatomy, pathomechanics, and pathophysiology of lateral ankle instability. J Athl Train 2002;37(4):364-75.
296 297
5. Hiller CE, Kilbreath SL, Refshauge KM. Chronic ankle instability: Evolution of the model. Journal of Athletic Training 2011;46(2):133-41.
298 299
6. Delahunt E, Coughlan GF, Caulfield B, Nightingale EJ, Lin CW, Hiller CE. Inclusion criteria when investigating insufficiencies in chronic ankle instability. Med Sci Sports Exerc 2010;42(11):2106-21.
300 301 302
7. Wright CJ, Arnold BL, Ross SE, Ketchum JM, Ericksen JJ, Pidcoe PE. Clinical exam results differ among individuals with functional ankle instability and ankle sprain copers. J Athl Train 2012;In Press.
303 304
8. de Noronha M, Refshauge KM, Crosbie J, Kilbreath SL. Relationship between functional ankle instability and postural control. J Orthop Sports Phys Ther 2008;38(12):782-9.
305 306
9. Arnold BL, Wright CJ, Ross SE. Functional ankle instability and health-related quality of life. J Athl Train 2011;46(6):634-41.
307 308 309
10. Shields CA, Needle AR, Rose WC, Swanik CB, Kaminski TW. Effect of elastic taping on postural control deficits in subjects with healthy ankles, copers, and individuals with functional ankle instability. Foot Ankle Int 2013;34(10):1427-35.
310 311 312
11. De Noronha M, Refshauge KM, Kilbreath SL, Figueiredo VG. Cross-cultural adaptation of the brazilian-portuguese version of the cumberland ankle instability tool (CAIT). Disabil Rehabil 2008;30(26):1959-65.
313 314 315
12. Cruz-Díaz D, Hita-Contreras F, Lomas-Vega R, Osuna-Pérez MC, Martínez-Amat A. Cross-cultural adaptation and validation of the spanish version of the cumberland ankle instability tool (CAIT): An instrument to assess unilateral chronic ankle instability. Clin Rheumatol 2013;32(1):91-8.
316 317
13. Donahue M, Simon J, Docherty CL. Critical review of self-reported functional ankle instability measures. Foot Ankle Int 2011;32(12):1140-6.
318 319
14. Hiller CE, Refshauge KM, Bundy AC, Herbert RD, Kilbreath SL. The cumberland ankle instability tool: A report of validity and reliability testing. Arch Phys Med Rehabil 2006;87(9):1235-41.
320 321
15. Hiller CE, Refshauge KM, Herbert RD, Kilbreath SL. Balance and recovery from a perturbation are impaired in people with functional ankle instability. Clin J Sport Med 2007;17(4):269-75.
AC C
EP
TE D
M AN U
SC
RI PT
287
14
ACCEPTED MANUSCRIPT
16. de Noronha M, Refshauge KM, Kilbreath SL, Crosbie J. Loss of proprioception or motor control is not related to functional ankle instability: An observational study. Aust J Physiother 2007;53(3):193-8.
324 325 326 327
17. Gribble PA, Delahunt E, Bleakley C, Caulfield B, Docherty CL, Fourchet F, Fong D, Hertel J, Hiller C, Kaminski TW, et al. Selection criteria for patients with chronic ankle instability in controlled research: A position statement of the international ankle consortium. J Orthop Sports PhyS Ther 2013;43(8):585-91.
328 329 330
18. Brown C, Padua D, Marshall SW, Guskiewicz K. Individuals with mechanical ankle instability exhibit different motion patterns than those with functional ankle instability and ankle sprain copers. Clin Biomech 2008;23(6):822-31.
331 332
19. Hubbard TJ. Ligament laxity following inversion injury with and without chronic ankle instability. Foot Ankle Int 2008;29(3):305-11.
333 334 335
20. Wikstrom EA, Tillman MD, Chmielewski TL, Cauraugh JH, Naugle KE, Borsa PA. Self-assessed disability and functional performance in individuals with and without ankle instability: A case control study. J Orthop Sports Phys Ther 2009;39(6):458-67.
336 337
21. Wikstrom EA, Fournier KA, McKeon PO. Postural control differs between those with and without chronic ankle instability. Gait Posture 2010;32(1):82-6.
338 339 340
22. Wikstrom EA, Tillman MD, Chmielewski TL, Cauraugh JH, Naugle KE, Borsa PA. Dynamic postural control but not mechanical stability differs among those with and without chronic ankle instability. Scand J Med Sci Sports 2010;20(1):e137-44.
341 342
23. Brown C. Foot clearance in walking and running in individuals with ankle instability. Am J Sports Med 2011;39(8):1769-76.
343 344
24. Wikstrom, E.A., Hass, C.J. Gait termination strategies differ between those with and without ankle instability. Clin Biomech 2012;27(6):619-24.
345 346 347
25. Croy T, Saliba SA, Saliba E, Anderson MW, Hertel J. Differences in lateral ankle laxity measured via stress ultrasonography in individuals with chronic ankle instability, ankle sprain copers, and healthy individuals. J Orthop Sports Phys Ther 2012;42(7):593-600.
348 349
26. Wikstrom EA, Tillman MD, Chmielewski TL, Cauraugh JH, Naugle KE, Borsa PA. Discriminating between copers and people with chronic ankle instability. J Athl Train 2012;47(2):136-42.
350
27. Youden WJ. Index for rating diagnostic tests. Cancer 1950;3(1):32-5.
351 352 353
28. Jaeschke R, Guyatt GH, Sackett DL. Users' guides to the medical literature. III. how to use an article about a diagnostic test. B. what are the results and will they help me in caring for my patients? the evidence-based medicine working group. JAMA 1994;271(9):703-7.
354 355
29. Freeman MR, Dean ME, Hanham IF. The etiology and prevention of functional instability of the foot. J Bone Joint Surg Br 1965;47(4):678-85.
356 357
30. Arnold BL, Linens SW, de la Motte SJ, Ross SE. Concentric evertor strength differences are associated with functional ankle instability: A meta-analysis. J Athl Train 2009;44(6):653-62.
AC C
EP
TE D
M AN U
SC
RI PT
322 323
15
ACCEPTED MANUSCRIPT
EP
TE D
M AN U
SC
RI PT
31. Hiller CE, Nightingale EJ, Raymond J, Kilbreath SL, Burns J, Black DA, Refshauge KM. Prevalence and impact of chronic musculoskeletal ankle disorders in the community. Arch Phys Med Rehabil 2012;93(10):1801-7.
AC C
358 359 360
16
ACCEPTED MANUSCRIPT
361
LEGEND TO FIGURES
362 Figure 1. Receiver operating characteristic (ROC) curve for Dataset 1 Cumberland Ankle Instability Tool
364
(CAIT) scores. Solid line = ROC curve, Doted line = reference line for significant ROC curve. *Cutoff value
365
with highest Youden Index
RI PT
363
366
Figure 2. Receiver operating characteristic (ROC) curve for Dataset 2 Cumberland Ankle Instability Tool
368
(CAIT) scores. Solid line = ROC curve, Doted line = reference line for significant ROC curve. *Cutoff value
369
with highest Youden Index in Dataset 2. Cutoff value identified in Dataset 1.
SC
367
†
AC C
EP
TE D
M AN U
370
17
ACCEPTED MANUSCRIPT
371
LEGEND TO TABLES
372 373
Table 1. Subject Demographics Dataset 1
375
Table 2. Subject Demographics Dataset 2
376 377
Table 3. Clinimetric properties at each potential CAIT cutoff using Dataset 1
SC
378
RI PT
374
Table 4. Clinimetric properties at each potential CAIT cutoff using Dataset 2
380
Table 5. Comparison of CAIT clinimetric properties between datasets
AC C
EP
TE D
M AN U
379
18
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Table 1. Subject Demographics Dataset 1 Descriptor Uninjured t-statistic df P-value CAI Gender 26 males 20 males, 35 females 37 females Age, yrs 25.52 ± 6.31 25.02 ± 5.49 -0.464 116 0.643 Height, m 1.71 ± 0.08 1.69 ± 0.08 -1.057 116 0.293 Weight, kg 77.07 ± 16.02 69.16 ± 13.06 -2.931 116 0.004* Abbreviations: CAI = Chronic Ankle Instability, CAIT = Cumberland Ankle Instability Tool. Values are mean ± standard deviation. *Significant difference in weight between CAI and uninjured subjects
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Table 2. Subject Demographics Dataset 2 Descriptor Coper Uninjured F-statistic P-Value CAI Gender 15 males, 12 males, 14 males, 14 females 14 females 13 females Age, yrs 23.31 ± 3.53 23.35 ± 3.50 22.89 ± 3.78 F2,79=0.135 0.874 0.847 Height, m 1.72 ± 0.10 1.71 ± 0.07 1.71 ± 0.08 F2,79=0.166 Weight, kg 75.12 ± 19.52 69.81 ± 13.65 69.33 ± 13.90 F2,79=1.128 0.329 Abbreviations: CAI = Chronic Ankle Instability, CAIT = Cumberland Ankle Instability Tool. Values are mean ± standard deviation.
ACCEPTED MANUSCRIPT
RI PT
LR1.000 0.984 0.967 0.951 0.934 0.902 0.836 0.754 0.672 0.623 0.525 0.410 0.344 0.246 0.180 0.082 0.049 0.051 0.017 0.000 0.000 -
TE D
M AN U
SC
Table 3. Clinimetric properties at each potential CAIT cutoff using Dataset 1. CAIT Sn 1-Sp Sp Youden Index LR+ 7.0 0.000 0.000 1.000 0.000 8.5 0.016 0.000 1.000 0.016 9.5 0.033 0.000 1.000 0.033 11.0 0.049 0.000 1.000 0.049 13.0 0.066 0.000 1.000 0.066 14.5 0.098 0.000 1.000 0.098 15.5 0.164 0.000 1.000 0.164 16.5 0.246 0.000 1.000 0.246 17.5 0.328 0.000 1.000 0.328 18.5 0.377 0.000 1.000 0.377 19.5 0.475 0.000 1.000 0.475 20.5 0.590 0.000 1.000 0.590 21.5 0.656 0.000 1.000 0.656 22.5 0.754 0.000 1.000 0.754 23.5 0.820 0.000 1.000 0.820 24.5 0.918 0.000 1.000 0.918 25.5 0.951 0.000 1.000 0.951 26.5 0.951 0.035 0.965 0.916 27.171 27.5 0.984 0.070 0.930 0.914 14.057 28.5 1.000 0.158 0.842 0.842 6.329 29.5 1.000 0.211 0.789 0.789 4.739 31.0 1.000 1.000 0.000 0.000 1.000
AC C
EP
Abbreviations: CAIT = Cumberland Ankle Instability Tool, Sn = sensitivity, Sp = Specificity, LR+ = Positive likelihood ratio, LR- = Negative Likelihood ratio
ACCEPTED MANUSCRIPT
RI PT
LR1.000 0.966 0.931 0.897 0.793 0.759 0.690 0.621 0.517 0.345 0.241 0.072 0.073 0.039 0.000 0.000 0.000 0.000 -
M AN U
SC
Table 4. Clinimetric properties at each potential CAIT cutoff using Dataset 2. CAIT Sn 1-Sp Sp Youden Index LR+ 5.0 0.000 0.000 1.000 0.000 9.0 0.034 0.000 1.000 0.034 13.0 0.069 0.000 1.000 0.069 15.0 0.103 0.000 1.000 0.103 16.5 0.207 0.000 1.000 0.207 17.5 0.241 0.000 1.000 0.241 18.5 0.310 0.000 1.000 0.310 19.5 0.379 0.000 1.000 0.379 20.5 0.483 0.000 1.000 0.483 21.5 0.655 0.000 1.000 0.655 22.5 0.759 0.000 1.000 0.759 23.5 0.931 0.038 0.962 0.893 24.500 24.5 0.931 0.057 0.943 0.874 16.333 25.5 0.966 0.132 0.868 0.834 7.318 26.5 1.000 0.189 0.811 0.811 5.291 27.5 1.000 0.458 0.542 0.542 2.183 28.5 1.000 0.472 0.528 0.528 2.119 29.5 1.000 0.623 0.377 0.377 1.605 31.00 1.000 1.000 0.000 0.000 1.000
AC C
EP
TE D
Abbreviations: CAIT = Cumberland Ankle Instability Tool, Sn = sensitivity, Sp = Specificity, LR+ = Positive likelihood ratio, LR- = Negative Likelihood ratio
ACCEPTED MANUSCRIPT
Table 5. Comparison of CAIT clinimetric properties between datasets Hiller 2006 Dataset 1 Test Characteristic (cutoff: ≤27) (cutoff: ≤25)
Dataset 2 (cutoff: ≤25)
AC C
EP
TE D
M AN U
SC
RI PT
Sensitivity 0.860 0.951 0.966 Specificity 0.830 1.000 0.868 a Positive Likelihood ratio 4.890 27.171 7.318 Negative Likelihood ratio 0.180 0.051 0.039 Abbreviations: CAIT = Cumberland Ankle Instability Tool a Due to perfect specificity, the positive likelihood ratio could not be calculated at the ≤25 cutoff, the value given is for the next closest value (≤26) where it could be calculated.
ACCEPTED MANUSCRIPT
M AN U
SC
RI PT
25.5*
AC C
EP
TE D
Figure 1.
ACCEPTED MANUSCRIPT
25.5†
M AN U
SC
RI PT
23.5*
AC C
EP
TE D
Figure 2.