The Ability of Vestibular and Oculomotor Screenings to Predict Recovery in Patients After Concussion: A Systematic Review of the Literature

This systematic review was completed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.²²  This review was registered in PROSPERO (no. CRD42022306344). The databases searched include PubMed, Ovid MEDLINE, SPORTDiscus, and the Cochrane Central Register of Controlled Trials (CENTRAL). The search terms used include VOMS, concussion, recovery, vestibular, vision, saccades, NPC, visual motion sensitivity, and King-Devick. The specific searches were concussion AND VOMS AND recovery, concussion and King-Devick AND recovery, and concussion AND vestibular OR vision OR saccades OR near point convergence OR visual motion sensitivity. In addition to the searches of databases, hand searches of reference lists of all relevant articles were conducted. The initial searches were conducted in January 2022. Follow up searches of PubMed were conducted in July 2022.

The inclusion criteria were studies of humans in which researchers assessed patients with the VOMS or another vestibular or oculomotor screening, recorded duration of recovery from concussion, English language, limited to the past 10 years (2012 to 2022), and level 4 evidence or higher based on the Centre for Evidence-Based Medicine scale.²³  Exclusion criteria were studies in which researchers examined rehabilitation effects on recovery. The results from the database searches were exported and inputted into Covidence systematic reviewer software (Veritas Health Innovation). Once the results were imported, 2 authors (M.B. and T.V.M.) screened the titles, abstracts, and full text for inclusion. The Mixed-Methods Appraisal Tool²⁴  was used to assess the quality of studies and was performed within Covidence (Veritas Health Innovation).

The primary data extracted from the studies included recovery time, vestibular or oculomotor scores, and other objective or subjective measures taken by clinicians or reported by patients, such as postconcussion symptom score, patient-report outcome measures, and cognitive or balance assessments. Secondary data extracted from the studies included the number of patients, patient demographics, study design, start and end date, inclusion and exclusion criteria, and method of recruitment. The data extraction was performed through Covidence and was organized by individual study. Each of the articles was evaluated, and the data were extracted and inputted into the individual data extraction tables in Covidence. If it was impractical to extract the data from within Covidence, such as that embedded in a large table, an Excel spreadsheet was used.

Once data were extracted, the authors critically evaluated each of the articles for commonalities and differences, focusing specifically on the extracted variables. The data from included studies were reviewed for a possible meta-analysis; however, the methods and statistical analyses used in individual studies did not allow for the extraction of similar data to perform a meta-analysis. A qualitative synthesis of the findings was used to summarize the conclusions. Articles were categorized into 2 groups based on the study design and comparators used. The groups included studies comparing typical and prolonged recovery and studies assessing recovery over time with a positive or negative initial vestibular or oculomotor screening finding.

Figure 1 presents the flow diagram based on the PRISMA guidelines. The literature search retrieved 1612 articles published between January 1, 2012, and January 2, 2022. We removed 684 duplicates and screened 937 studies. After we screened titles and abstracts, we assessed 42 articles for inclusion based on their full text. Two articles were included in the full text screening after a review of the author’s personal article databases. After a full text evaluation, we included 18 articles in the review (Tables 1–3).

The Mixed-Methods Appraisal Tool was used to evaluate each study for its quality (Tables 4–6). All the included studies had representative samples of their respective populations of interest and used appropriate measurements. Most of the included studies had complete outcome data, except 4 studies.^25–28  Although most studies had complete outcome data, a few had adequate controls for confounding variables.^{26,27,29–34}  Finally, for all of the studies, exposures occurred as intended with the exception of one study²⁶  in which this aspect could not be determined.

In all of the included studies, the authors conducted vestibular or oculomotor screening assessments at either the time of injury or during the initial clinic encounter. Researchers of some studies included subsequent screenings at follow-up visits. For our purposes we did not operationalize a positive or negative vestibular or oculomotor screening assessment. However, several studies (Büttner et al,³⁵  Eagle et al,³⁰  Glendon et al,³⁶  Henry et al,²⁵  Price et al,¹¹  Sinnott et al,³⁷  Whitney et al,³⁸  and Worts et al³⁹ ) used the VOMS cutoff of greater than a 2-symptom score increase described by Mucha et al.²¹  Other authors (Anzalone et al,⁴⁰  Leddy et al,²⁶  Master et al,⁴¹  and Martinez et al³⁴ ) chose to dichotomize patients based on any symptom provocation. Cheever et al,²⁸  Glendon et al,³⁶  and Sufrinko et al³²  did not disclose which they used, and Knell et al³¹  examined different cut points for the VOMS.

Researchers of nine of the included studies dichotomized patient populations into typical and prolonged recovery. The most common cutoff for typical vs prolonged recovery was less than or equal to 28 days and greater than 28 days used in 3 studies.^29,33,39  Martinez et al³⁴  used less than 29 days and greater than or equal to 29 days; Price et al,¹¹  Ellis et al,²⁷  and Eagle et al³⁰  used less than or equal to 30 days and greater than 30 days as their recovery cutoff; Leddy et al²⁶  stratified at 21 days; Cheever et al²⁸  stratified at 16 days; and Sufrinko et al³²  stratified patients into groups of less than 15 days, 15 to 29 days, and greater than 29 days until recovery. For our purposes, we used the stratification by Sufrinko et al³² ; we combined cutoff groups of less than 15 days and 15 to 29 days into one group of less than or equal to 29 days.

Among studies that dichotomized recovery, researchers of only 3 studies (Leddy et al,²⁶  Sufrinko et al,³²  and Worts et al³⁹ ) investigated the association between a positive test of smooth pursuits at initial presentation and recovery. In 2 studies, the authors found a significant relationship between positive smooth pursuits at initial presentation and prolonged recovery, with Worts et al noting higher symptom provocation from smooth pursuits among the prolonged patients (P = .008)³⁹  and Sufrinko et al finding smooth pursuits had the strongest association with recovery over 30 days (odds ratio [OR] = 1.5 [95% CI = 1.19, 1.90], P < .001).³²  Although Leddy et al²⁶  did not find a statistically significant relationship between a positive smooth pursuit at initial examination (1 to 10 days postinjury), upon a second visit (13.6 ± 1 days after visit 1) the authors did find a statistically significant association (P = .0001) between a positive smooth pursuit test and longer recovery at the second visit. When comparing saccadic testing and prolonged recovery, Worts et al³⁹  found that patients who had a higher symptom provocation for either horizontal or vertical saccades were more likely to be in the prolonged recovery group (horizontal saccades: P = .003, crude (unadjusted) OR = 2.087 [95% CI = 1.375, 3.167]; vertical saccades: P = .039, crude OR = 1.462 [95% CI = 0.983, 2.173]). Similarly, Sufrinko et al³²  found that a positive smooth pursuit was predictive of recovery of 30 to 90 days (OR = 1.5 [95% CI = 1.19, 1.90], P < .001) and a positive smooth pursuit, horizontal or vertical saccades was predictive of a recovery of 15 to 29 days (smooth pursuits: OR = 1.25 [95% CI = 1.02, 1.55], P = .036; horizontal saccades: OR = 1.31 [95% CI = 1.06, 1.62], P = .013; vertical saccade: OR = 1.22 [95% CI = 1.01, 1.47], P = .035).

A similar assessment of saccadic functioning, the King-Devick test, was included in 4 studies within this subgroup that dichotomized recovery.^11,28,34,39  Among these studies, only Worts et al³⁹  found that patients who took longer to complete the King-Devick test were more likely to have a prolonged recovery (total time: P < .001; crude OR = 2.984 [95% CI = 1.958, 4.546]). Specifically, patients in the prolonged group took a longer (73.2 ± 21.7 seconds) time to complete the King-Devick test than those in normal recovery (56.9 ± 14.0 seconds).³⁹  Martinez et al³⁴  found that patients who had symptom provocation during the test were almost 3 times as likely to have a prolonged recovery (OR = 2.98, P < .01). Price et al¹¹  found a small but nonsignificant effect of 1.09 (95% CI = 0.88, 1.34), demonstrating a potential increased likelihood of prolonged recovery when testing took longer than the 75th percentile for the respective age group on the King-Devick test. In contrast, Cheever et al²⁸  found no statistically significant difference between performance on the King-Devick test between patients who had a prolonged vs normal recovery at initial visit or any subsequent visit (initial visit: P = .214; 2 weeks: P = .779; 6 weeks: P = .698).

A positive NPC appears to be associated with prolonged concussion recovery, according to 4 studies,^26,29,33,39  which reported statistically significant relationships. Sufrinko et al³²  found a similar result comparing convergence distance predictive of a recovery of 15 to 90 days compared to a recovery of ≤14 days. Cheever et al²⁸  did not find an association between a positive NPC and the likelihood of prolonged recovery at initial visit (P = .114); however, the authors used a shorter elapsed time to dichotomize patients into prolonged and normal recovery, namely, 16 days.

Researchers of four studies^26,30,32,39  examined VOR testing with mixed results. Eagle et al³⁰  reported a significant relationship between a positive VOR and longer recovery time in both male and female patients. Sufrinko et al³²  found positive horizontal and vertical VOR predictive of longer recovery with reported ORs as high as 1.31 by using the horizontal VOR test to predict a recovery of 30 to 90 days compared to a recovery less than 14 days. Interestingly, in the same study, the horizontal and vertical VOR tests were unable to predict a recovery of 15 to 29 days compared to a recovery of less than 14 days when using multinomial regressions.³²  Counter to these results, researchers of 2 studies (Leddy et al²⁶  and Worts et al³⁹ ) found no statistically significant differences between patients with a prolonged and normal recovery when comparing results of VOR testing. Worts et al³⁹  did find numerical differences between the prolonged and normal groups on VOR testing, but the differences did not reach statistical significance.

Eagle et al,³⁰  Sufrinko et al,³²  and Worts et al³⁹  examined visual motion sensitivity (VMS) with similar findings across studies that a positive VMS is related to prolonged recovery. Eagle et al³⁰  found a significant relationship between a longer recovery and positive VMS. When Worts et al³⁹  included extreme test scores in regression models, they found that VMS is likely to predict prolonged recovery with an OR of 1.54 (95% CI = 1.13, 2.09) and a crude OR of 4.36 (95% CI = 2.13, 8.93) for an adjusted OR. Sufrinko et al³²  found in a multinomial regression analysis that a patient with a positive VMS was 1.23 times as likely to have a 30- to 90-day recovery relative to a recovery less than or equal to 14 days.

Finally, researchers of 5 studies^{11,27,30,34,39}  reported on grouped vestibular and oculomotor testing positivity and its relationship to prolonged recovery. Worts et al³⁹  found all of the VOMS measures, except VOR, to be predictive of prolonged recovery. Price et al¹¹  reported a negative VOMS predictive of normal recovery (fully adjusted incidence rate ratio = 1.31 [95% CI = 1.05, 1.63), with a negative predictive value of 80.77%. Eagle et al³⁰  found that any increase in symptoms during VOMS testing was predictive of a longer recovery in males only, and in females, only the vestibular assessments within the VOMS was predicative of prolonged recovery. Similarly, if a patient did not have a 1-unit increase, researchers found a 90% chance that patients recovered within 30 days.³⁰  Martinez et al³⁴  found that patients were 4 times more likely to have a prolonged recovery if they had symptom provocation on any of the screening tests (OR = 4.28, P < .01). Additionally, the same study found consistently, across multiple analyses, that a positive vestibular screen was associated with a 3 to 4 times increased odds of delayed recovery in patients after concussion.³⁴  Ellis et al²⁷  noted that more than 1 subjective report of vision problems and 1 positive oculomotor or vestibular test were predictive of prolonged recovery (adjusted OR = 4.1 [95% CI = 1.04, 16.16]). A positive oculomotor or vestibular test in the previous study²⁷  included symptom provocation or an abnormal finding on a singular test. Researchers of these 5 studies demonstrated consistently that the VOMS, vestibular screenings, or oculomotor screenings may help predict recovery.

Although many study designs classified patients into normal and prolonged recovery, others chose to use recovery as a continuous variable. Researchers of four studies^37,38,40,41  used continuous days until recovery or return to play. Anzalone et al,⁴⁰  Master et al,⁴¹  and Whitney et al³⁸  reported on smooth pursuits and its relationship to recovery; all researchers found a significantly longer recovery in patients who had a positive smooth pursuits test than patients who had a negative test. Similarly, a positive saccades was found to be significantly associated with a longer recovery by Whitney et al³⁸  (P = .025) and Anzalone et al⁴⁰  (horizontal saccades: P = .018, vertical saccades: P < .001), a positive VOR was found by Anzalone et al⁴⁰  (horizontal VOR: P = .018, vertical VOR: P = .002) and Master et al⁴¹  (log-rank P = .001), and a positive NPC was found by Whitney et al³⁸  (P = .031).

Whitney et al³⁸  found that patients who had a VOMS symptom increase of greater than or equal to 2 had an average recovery time of 3.5 days longer than those who had a negative VOMS. Knell et al³¹  assessed symptom increase within sexes and found that (1) in males, a 1-unit increase in symptom provocation during VOMS assessment increased recovery time by 1.38 days (P < .001), and (2) in females, the recovery duration per 1-unit increase was 1.73 days (P < .001). Similarly, Price et al¹¹  reported that patients who had a 2-point increase on any VOMS test had a 1.31-day longer recovery than patients with a lesser increase. Researchers of another study found that patients who reported more than 1 vision problem and at least 1 positive oculomotor or vestibular test had a median recovery time of 40 days (interquartile range = 28.5–54 days) compared to 21 days (interquartile range = 13–32 days) in patients who did not report these symptoms and had negative oculomotor or vestibular tests.²⁷ 

Additionally, Sinnott et al³⁷  examined recovery among patients who had no vestibular symptoms compared to those who reported their vestibular symptoms were improving over time and patients who indicated their vestibular symptoms were persisting. In patients with improving vestibular symptoms, recovery was 3.47 days longer than patients with no vestibular symptoms. The difference reported by Sinnott et al³⁷  between patients who had no symptoms and persisting symptoms was even greater with patients experiencing persisting symptoms having a 11.99-day longer recovery than patients with no vestibular symptoms. Researchers of these studies consistently suggest that a longer recovery is associated with positive vestibular or oculomotor tests.

Researchers of 14 of the 19 included studies in this systematic review suggest that patients having a positive vestibular or oculomotor screening test is indicative of longer recovery than patients who do not test positive on these screenings. Authors of these studies consistently found that vestibular or oculomotor tests are able to predict recovery, with authors of studies reporting a difference of slightly more than 1 day.¹¹  In contrast, other authors report a 3- to 4-day difference in recovery length.^31,37,38  It appears that if patients report vestibular or oculomotor symptoms or screen positive on vestibular or oculomotor testing, they will have a longer recovery than patients who do not because these systems affect how people interact with their environment. Other researchers have found the VOMS to have relatively low false-positive rates⁴²  and similar impairments between collegiate and high school athletes.⁴³  Moreover, sideline assessments of oculomotor testing, like the King-Devick test, have been found to be a good supplemental tool to traditional sideline assessment tools, such as the SCAT5.⁴⁴  Within the broader prognostic indicator literature for concussion, this review builds upon previous literature^12,13  in which authors suggest that vestibular or oculomotor dysfunction may play a role in identifying patients with persisting symptoms who have a longer recovery. As such, the vestibular and oculomotor assessments should be included as part of a multimodal concussion assessment battery.

Our findings should be interpreted in the context of several limitations. Most patients were seen in concussion specialty care; the aggregation of findings suffered from inconsistent methods, including assessment tool and timing; and outcome measures were inconsistent. Within the studies included in the systematic review in which researchers examined VOMS testing, 8 studies (Büttner et al,³⁵  Eagle et al,³⁰  Glendon et al,³⁶  Henry et al,²⁵  Price et al,¹¹  Sinnott et al,³⁷  Whitney et al,³⁸  and Worts et al³⁹ ) classified a positive VOMS as symptom provocation of greater than 2, as proposed in the original article for VOMS testing by Mucha et al.²¹  Authors of 4 articles (Anzalone et al,⁴⁰  Leddy et al,²⁶  Master et al,⁴¹  and Martinez et al³⁴ ) used any symptom provocation as a positive VOMS. This discrepancy makes it difficult to compare studies because of the different operationalizations for which some patients in the latter group of studies would test positive. In the former group, if the symptoms did not reach a threshold greater than 2, the patients would have been noted as negative. The populations studied in the included articles ranged from as young as 5 years old to 40 years old. However, only Büttner et al³⁵  included athletes who were older than collegiate athletes, and authors of 2 studies (Ellis et al²⁷  and Walker et al³³ ) did not exclude patients younger than 5 years old in their methods. The weighted mean age for studies that reported mean age is 16.4 years old, demonstrating that the patient population is concentrated in high school-aged patients. The adolescent age group appears to sustain an inordinate number of concussions, as supported by epidemiologic research.⁴⁵  Having most patients treated in concussion specialty care limits the clinical applicability of this review. Patients who are seen in specialty care often have access to private insurance and are of a higher socioeconomic status, and thus, they are able to seek care earlier, which has also been found to reduce recovery time.⁴⁶  Also, our findings may not be generalizable to patients in lower socioeconomic situations. Additionally, patients evaluated in specialty clinics may have had symptoms longer⁴⁷  or more severe symptoms than patients who were evaluated by only an athletic trainer or primary care provider. These patients may not represent the larger proportion of patients whose concussions do not require specialty care.

Additionally, the methods of the included studies were varied, and the authors were unable to combine the results of these studies into a meta-analysis. Furthermore, the authors reported a variety of evaluation timepoints, outcomes, and analytical approaches, including VOMS component interpretation, which limited the direct comparison of results across studies (Supplemental Table). Furthermore, recent evidence suggests that using a change score with the VOMS, rather than the total score, may improve the identification of concussion through component and overall clinical cut-offs.⁴⁸  Future researchers should use multiple variables in the concussion evaluation to help clinicians predict a prolonged recovery while using the shortest clinical examination.

Vestibular and oculomotor screening tests can help inform clinicians about a patient’s status. With the findings of this systematic review, we conclude that a positive vestibular or oculomotor assessment may be prognostic of a longer recovery time than a negative screening (Strength of Recommendation: B). Although researchers of most studies found that a positive vestibular or oculomotor screen was associated with a longer recovery, most of the included studies were level 3 or 4 evidence and are not generalizable to patient populations. However, clinicians in all settings should be encouraged to include vestibular or oculomotor assessments into their concussion assessment battery to inform rehabilitative efforts and monitor recovery (Strength of Recommendation: C). These assessments are cost effective, require little to no equipment, and can assist clinicians in determining dysfunction, informing rehabilitative efforts, and aiding in determining a patient’s prognosis.