Receiver operating characteristic curve analysis of clinical signs for screening of convergence insufficiency in young adults

Convergence insufficiency (CI) is a dysfunction of binocular vision that is associated with various signs and symptoms in near work. However, CI screening is performed less frequently in adults than in children. We aimed to evaluate the ability of screening tests to discriminate CI from other binocular vision anomalies and normal binocular vision in young adults. One hundred eighty-four university students (age, 18–28 years) who underwent an eye examination due to ocular discomfort were included. Near point of convergence (NPC), phoria, accommodative amplitude, fusional vergence, the ratio of accommodative convergence to accommodation, relative accommodation, binocular accommodative facility, vergence facility, and the values corresponding to Sheard’s and Percival’s criteria were evaluated. Receiver operating characteristic (ROC) curve analysis for each test was also performed. The prevalence of CI ranged from 10.3% to 21.2%, depending on the signs and the presence of CI associated with accommodative disorders. Assessments based on NPC, Sheard’s criterion, and Percival’s criterion showed high discriminative ability, with the ability being higher between the CI and normal binocular vision groups than between the CI and non-CI groups. Sheard’s criterion showed the highest diagnostic performance in discriminating CI with three signs from the non-CI group. The cut-off values were 7.2 cm for NPC, -0.23 to 1.00 for Sheard’s criterion, and -4.00 to -2.33 for Percival’s criterion. Our results suggest that the use of Sheard’s criterion with NPC shows high performance for screening of CI.

If the data are held or will be held in a public repository, include URLs, accession numbers or DOIs. If this information will only be available after acceptance, indicate this by ticking the box below. For example: All XXX files are available from the XXX database (accession number(s) XXX, XXX.).
• If the data are all contained within the manuscript and/or Supporting Information files, enter the following: All relevant data are within the manuscript and its Supporting Information files.
• If neither of these applies but you are able to provide details of access elsewhere, with or without limitations, please do so. For example: Data cannot be shared publicly because of [XXX]. Data are available from the XXX Institutional Data Access / Ethics Committee (contact via XXX) for researchers who meet the criteria for access to confidential data.
The data underlying the results presented in the study are available from (include the name of the third party • All relevant data are within the manuscript and its Supporting Information files. and contact information or URL). This text is appropriate if the data are owned by a third party and authors do not have permission to share the data.

• * typeset
Additional data availability information: Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation deficiencies in negative relative accommodation (NRA) [1]. However, cases with CI 48 can be simple or complex aspects because not all patients with CI have all these 49 symptoms and signs [1,4]. The reported prevalence of CI varies widely from 1.0% to 50 33% [1,[5][6][7]. These wide variations can be attributed to the differences in patient age, 51 sample populations, and diagnostic criteria used in the previous studies [8]. Therefore, 52 it is important to evaluate the ability of clinical diagnostic tests to discriminate 53 between CI and other disorders, since the clinical criteria used for screening CI have 54 been different across several studies. 55 The area under the curve (AUC) in the ROC curve analysis indicates the 225 discriminative ability to distinguish between subjects with and without CI. The cut-off 226 value for each test was defined as the coordinate that had the maximized sum of 227 sensitivity and specificity. The cut-off was also identified for each test with the largest 228 AUC which was significantly greater than 0. 50 and Percival's criterion with statistically significant differences when comparing the 282 AUC to the value 0.5. Among the AUCs for all diagnostic tests included in Table 1, 283 the values were significantly greater than 0.5 for only the NPC and NFV at distance, 284 and the AUCs for diagnostic tests excluded in Table 1 were also significantly greater 285 than 0.5 for Sheard's and Percival's criteria. AUCs were greater for NBV than for 286 non-CI, greater for excluding than for including CI associated with AD, and greater 287 for CI with three signs than for CI with two signs.   Table 5 shows the sensitivity, specificity, positive likelihood ratio (LR+), and 301 negative likelihood ratio (LR-) for each test by using cut-offs obtained with ROC 302 curves. The NPC cut-offs were >7.2 cm for all classified CIs combined with the NBV 303 and non-CI groups. The Sheard's criterion cut-off of >1.00 for CI with three signs 304 (CI3) was higher than the cut-off of >-0.23 for CI with two signs (CI2), and showed 305 higher sensitivity and specificity than CI2. The Percival criterion cut-off of >-4.00 for 306 Sheard's criterion in other studies [22,32], and these tests was limited to school-age 364 children. However, Percival's criterion as well as Sheard's criterion was applied to 365 adults in this study. There are significant differences between the all-CI and NBV 366 groups for Sheard's criterion at distance/near and Percival's criterion at near and 367 between the all-CI and other-BVA groups for Sheard's criterion at near. These results 368 indicate that Sheard's criterion could be used in tests to distinguish CI from other 369 BVA and NBV, and Percival's criterion could be used to distinguish CI from NBV. 370 However, Sheard's criterion is a useful tool for screening CI with exophoria 371 associated with near tasks because the signs of CI include exophoria more than 6 ∆ at 372 near and normal phoria of 0-6 ∆ exophoria at distance [17], and previous studies [33, 373 34] have suggested that Sheard's and Percival's criteria are the most effective with 374 exophoria and esophoria, respectively. 375 The main finding in this study was that NPC can distinguish individuals with CI 376 signs or CI signs associated with AD, namely, CI2AD and CI3AD from the NBV and 377 non-CI groups. In ROC curve analysis, the AUC of 0.842-0.920 obtained using the 378 NPC test represents an excellent discriminative ability for CI screening. Although the 379 test parameters such as subjects and diagnostic and classification criteria were not indicates a decreased probability that the NPC test is negative. 393 The ROC curve analysis in this study showed that Sheard's and Percival's criteria 394 have potential for use as tools for CI screening. Sheard's criterion is particularly 395 useful for CI screening from non-CI than from NBV. In addition, Sheard's criterion 396 can be a better tool than NPC in cases of CI with three signs (CI3, CI3AD). The 397 previous studies [9, 11] did not provide cut-off values, sensitivity, specificity, AUC, 398 and other data from ROC curve analysis for Sheard's and Percival's criterion to 399 screen CI. The AUC of 0.773-0.912 obtained using Sheard's criterion in our study 400 represents an acceptable discriminative ability for CI screening, and the AUC reduced 401 with decreasing signs. Sheard's criterion in this study has positive cut-offs values 402 (failed to normal binocular vision or needed prism) in the CI3 and cut-offs of > -0.23 403 (approximately cut-offs > zero) in the CI2. Sheard's criterion could diagnose all-CI. 404 The sensitivity of the combined CI3 is higher than that of the combined CI2, and the 405 specificity is lower than sensitivity. Although the LR+ of 2.25-3.84 was lower than 406 the corresponding value for the NPC assessment as a positive test result indicating the 407 presence of CI, and the LR-of 0.22-0.28 was higher than the corresponding value for 408 the NPC assessment as a negative result test indicating the absence of CI, from 409 another perspective, this criterion is a valid tool for discriminating CI with three signs 410 (CI3, CI3AD) from non-CI because the AUC of Sheard's criterion was greater than 411 that in NPC and the LR-of zero in CI3 and CI3AD was equal to that for NPC. These differences could have occurred because Sheard's criterion works best for 422 exophoric conditions such as CI and Percival's criterion tends to work best for near 423 esophoric conditions such as convergence excess [23,37]. However, Percival's 424 criterion showed a lower discriminative ability than Sheard's criterion for CI 425 screening. In ROC curve analysis, Percival's criterion showed lower AUC and LR+ 426 than Sheard's criterion. Percival's criterion also showed lower sensitivity than 427 Sheard's criterion, except for CI2 combined with non-CI (CI2 + non-CI) and CI2 428 combined with AD (CI2AD + NBV, CI2AD + non-CI  showed high discriminative ability, with the ability being higher between the CI and 29 normal binocular vision groups than between the CI and non-CI groups. Sheard's 30 criterion showed the highest diagnostic performance in discriminating CI with 3 three 31 signs from the non-CI group. The cut-off values were 7.3 2 cm for NPC, -0.23 to 1.00 32 for Sheard's criterion, and -4.00 to -2.33 for Percival's criterion. Our results suggest 33 that the use of Sheard's criterion with NPC shows high performance for screening of 34 Our study was limited to young adults and an approach focused on signs and 91 modified signs such as Sheard's and Percival criterion, in addition to fusional 92 vergence related to ability to maintain single binocular vision. The purpose of this 93 study was to evaluate the accommodative and vergence ability for university students 94 who visited for primary eye care due to ocular discomfort and to analyze the 95 diagnostic ability of each test for screening CI by performing ROC curve analysis, 96 including assessments of sensitivity, specificity, cut-off values, and likelihood ratio. The participants were 184 university students (age, = 18-28 years; mean age, = 22.23 101 ± 2.26 years) who underwent an eye examination due to ocular discomfort. 102 Participating students voluntarily visited a university eye clinic center for primary eye 103 care due to blurred vision, eyestrain, and visual discomfort. This study was approved 104 by the Kangwon nNational uUniversity iInstitutional rReview bBoard (KWNUIRB-105 2019-02-001-002), with waiver of the informed consent for the retrospective 106 collection of clinical data, and adhered to the tenets of the Declaration of Helsinki. 107 Participants had not previously undergone any vison therapy or eye exercise 108 treatment. We excluded patients with ocular diseases such as glaucoma, cataract, and 109 retinal disease and those with a history of prior surgery, which was determined by 110 history-taking [1]. The criteria for inclusion into study also were the absence of 111 amblyopia and strabismus.   (BO) prism with the same method as NFVs. Fusional vergences at near were 161 measured after fusional vergence tests at distance [17]. If no blur value in fusional 162 8 vergence was reported, the break value was used in the analysis. The normative values 163 are 9 ± 2 ∆ for bur, 19 ± 4 ∆ for break, and 10 ± 2 ∆ for recovery of PFV at distance; 7 164 ± 2 ∆ for break, and 4 ± 1 ∆ for recovery of NFV at distance; 17 ± 3 ∆ for bur, 21 ± 3 165 ∆ for break, and 11 ± 4 ∆ for recovery of PFV at near; 13 ± 2 ∆ for bur, 21 ± 2 ∆ for 166 break, and 13 ± 3 ∆ for recovery of NFV at near. The relation of fusional vergence to phoria was analyzed by using Sheard's [2322] Table 1, which were modified from the study byreferred to 207 Scheiman and Wick's study [17] and compared with the expected criteria for each test 208 [2524]. Subjects groups were classified into three groups according to Table 1 to the probability that a test will indicate CI when CI is present, and specificity refers 229 to the probability that a test will indicate the absence of CI when CI is not present. 230 The area under the curve (AUC) in the ROC curve analysis indicates the 231 discriminative ability to distinguish between subjects with and without CI. The cut-off 232 value for each test was defined as the coordinate that had the maximized sum of 233 sensitivity and specificity. The cut-off was also identified for each test with the largest 234    Table 3 shows the prevalence of non-strabismic binocular vision anomalies 267 diagnosed according to Table 1

284
ROC curve analysis was performed for the tests shown in Table 2. The results of 285 the AUC for NPC, Sheard's and Percival's criterion with p < 0.05 and AUC > 0.5 in 286 95% confidence interval ranges, and PFV, NFV, and AC/A including the diagnostic 287 criteria in Table 1The AUC findings (p < 0.05 and AUC > 0.5 in 95% confidence 288 interval ranges) from ROC curve analyses for each CI test in various conditions are 289 shown in Table 4. to the value 0.5. Among the AUCs for all diagnostic tests included in Table 1, the 292 values were significantly greater than 0.5 for only the NPC and NFV at distance, and 293 the AUCs for diagnostic tests excluded in Table 1 were also significantly greater than 294 0.5 for Sheard's and Percival's criteria. AUCs were greater for NBV than for non-CI, 295 greater for excluding than for including CI associated with AD, and greater for CI 296 with 3 three signs than for CI with 2 two signs.  signs (CI3) was higher than the cut-off of >-0.23 for CI with 2 two signs (CI2), and 314 showed higher sensitivity and specificity than CI2. The Percival criterion cut-off of >-315 4.00 for CI2 was lower than the cut-off of >-2.33 for CI3. Sensitivity, specificity, and 316 LR+ were higher in the order of NPC, Sheard's, and Percival's criterion, but LR-317 showed the opposite trend.  In this study, with ROC curve analysis of signs, the tests that showed a significant 327 discriminative ability between CI and non-CI or CI and NBV among young adults 328 university students were those based on the Sheard's and Percival's criteria for near 329 vision, while NPC assessment was the best diagnostic test for identifying CI. For CI 330 screening between CI3 (CI with 3 three signs) and non-CI, Sheard's criterion was a 331 better diagnostic parameter than NPC. The distribution of CI, non-CI, and NBV 332 according to diagnostic criteria such as population and the signs of CI influenced the 333 validity of each test in the ROC curve analysis. 334 The prevalence of CI in this study ranged between 10.3% and 21.2%, higher than university students who underwent primary eye care due to ocular discomfort, and CI 342 was classified based on the basis of signs. The CI screening ability of each test was 343 also evaluated in conditions such as NBV and non-CI conditions. The variation in CI 344 prevalence in this study appears to be associated with differences in signs, 345 classification criteria, and population. In our study, the prevalence of myopia was 346 high (70.1%). CI has been reported to show a significant association with myopia 347 and Sheard's criterion in other studies [2322,3332], and these tests was limited to 374 school-age children. However, Percival's criterion as well as Sheard's criterion was 375 applied to adults in this study. There are significant differences between the all-CI and When submitting your revision, we need you to address these additional requirements.  Figure 1R below shows a part of the sample). Compared with other journals, the duplicated rate was 4% ( Figure 2R below also shows a part of the sample). We have rechecked and rephrased the detected text, and have summarized these revisions in Table 1R:   [17,24]. Normative values of each test have been included in the methods section. This 'which were modified from' is a describing error. 'which were modified from' have been replaced with 'which referred to'.

Results
Were all the subjects with a refractive error wearing their refractive correction for testing? R27R#2: 'after wearing refractive correction' has been added to lines 241-242 to clarify this.
211 -Here it is stated what the 3 groups are but this should have appeared in the method section. The groups are unclearly presented. What is meant by 'all CI'? Is this anyone that fell into the CI category in Table 1? Are those in 'BVA' anyone that falls into any other diagnosis based on Table 1? I assume NBV are those that have nothing wrong and it might be less confusing to refer to them as the control group. The abbreviation BVA is also confusing as that typically stands for binocular visual acuity R28R#2: In relation to the answer given in R26R#2, a description of the three groups has been noted to lines 203-208 of the method section as follows: 'Subjects groups were classified into three groups according to Table 1: All CI including CI with 2 signs and CI with an accommodative dysfunction, other binocular vision anomaly (BVA) except CI, and normal binocular vision (NBV). All CI were also classified into CI with 3 signs (CI3), CI with 2 signs and an accommodative dysfunction (CI3AD), and CI with 2 signs (CI2).' The abbreviation BVA was determined under our careful consideration. 'non-strabismic' was deleted to eliminate confusion in line 261 and from the title of Table 3. "normal group" has been replaced with "NBV" to avoid confusion in line 264. Table 3. R31R#2: The paragraph has been revised to use NBV, CI2, CI + AI, CI + AE, CI3, and CI2AD so that it matches with the terms used in Table 3. Table 3 -Again it is unclear who fits within the various groups. Does the 'non-CI' group contain everyone in the 'other BVA' group? Looking at the numbers it looks like this could be those in 'other BVA' and 'NBV'. Does AD include anyone of those listed with AE and AI? I would assume so but the numbers don't add up correctly. Why is FVD and accommodative infacility not included in this table? R32R#2: 'Non-CI corresponding to CI is represented in the non-CI group.' has been included in Table 3 to improve its clarity. All (including NBV) except classified CI (CI3, CI3AD, CI2, CI2AD) have been classified into the non-CI group. In the case of this study, FVD and accommodative infacility was not present.

232-38 Within this p/g indicate CI2AD etc so that it matches with the terms used in
246-53 -'for each CI test' is unclear. Firstly, it would be better to just refer to them as 'tests' rather than 'CI tests' and secondly, they are not all included in Table 4. Again it is stated 'for each test' yet Figure 1 only shows the ROC curves for NPC, Sheard and Percival. State why only particular tests are presented in Table 4 and figure 1. I assume 'for diagnostic tests excluded in Table 1' refers to Sheard and Percival? This should be explicitly stated, rather than requiring the reader to go back and work this out. R33R#2: 'ROC curve analysis was performed for the tests shown in Table 2. The results of the AUC for NPC, Sheard's and Percival's criterion with p < 0.05 and AUC > 0.5 in 95% confidence interval ranges, and PFV, NFV, and AC/A including the diagnostic criteria in Table 1 are shown in  Table 4 and figure 1' is that other tests except tests presented in Table 4 and Figure 1 were not significant and tests resented in Table 4 and Figure 1 were significant and high priority for CI screening ( Figure R1).

-'AUCs were greater for NBV than for non-CI'. In an earlier comment I had stated that it seemed that 'non-CI' included those in 'other BVA' and 'NBV', yet now it is suggested that this
is not the case. R34R#2: As in R30R#2 and R32R#2, when CI is classified as CI3, CI3AD, CI2, and CI2AD, respectively, the non-CI group includes all (other BVA and NBV) those except for the classified CI, respectively. For an example, Table 4 represents the ability to discriminate CI from CI3 + NBV, and discriminate CI from CI3 + non-CI.  Table 2, and 2) PFV, NFV, and AC/A including diagnostic criteria in Table 1. NFV for distance has been represented in a footnote in Table 4. 'screening for various CI conditions' has been revised to 'Ability to discriminate CI" to improve the clarity of Table 4. To further reduce confusion, the first column of Table 4 has also been revised. 301 -should state 'was also evaluated in NBV and non-CI conditions'? R37R#2: We have corrected the text to 'was also evaluated in NBV and non-CI conditions'.

311-2 -This sentence does not seem to link with the previous sentence.
R39R#2: This sentence has been corrected to 'CI should also be distinguished from conditions such as normal vision, diverse vergence and accommodative disorders, and other binocular anomalies.' 327unclear why discriminating CI from abnormal rather than normal groups. R40R#2: 'from abnormal groups combined with other binocular anomalies' has been added to lines 361-356 to clarify this.