Relationship between Static Ocular Counterroll and Bielschowsky Head Tilt Phenomenon

Abstract

Purpose.: This study was conducted to assess how hyperdeviation of a paretic eye during ipsilesional head tilt—the Bielschowsky head tilt phenomenon (BHP)—can be explained by decreased compensatory ocular counterrolling (OCR) due to the depressed torque of the paretic superior oblique (SO) muscle.

Methods.: Thirty-three patients with clinically diagnosed SO palsy and 11 control subjects were studied. With a head-mounted video camera, static ocular counterrolling (s-OCR) was determined by measuring the inclination of a line connecting the two centroids of the characteristic iris pattern and corneal reflex. The BHP was measured with the alternate prism and cover test.

Results.: The mean (SD) amplitude of s-OCR in paretic eyes based on the fit of the regression sine curve against the ipsilesional head tilt angle was significantly decreased compared with that for contralesional head tilt, 6.3 (3.5)° for ipsilesional and 11.3 (3.9)° for contralesional (P < 0.001), and was significantly smaller than that in normal subjects: 10.9 (2.6)° (P < 0.001). No significant linear relation was noted between hyperdeviation on ipsilesional head tilt and the amplitude of s-OCR in paretic eyes (r ² = 0.04; P = 0.29). However, the differences between the hyperdeviation with ipsilesional 30° head tilt and with head-upright position correlated significantly with the amplitudes of s-OCR in paretic eyes (r ² = 0.19, P = 0.01).

Conclusions.: The absolute value of the hypertropia on ipsilesional head tilt in clinically diagnosed SO palsy does not directly assess the function of the SO muscle. The difference in hypertropia between ipsilesional head tilt and the upright position, however, may be a better indicator of SO function.

When the head is tilted toward the shoulder, the two intorting muscles in the eye on the side to which the head is tilted are stimulated, and the two extorting muscles are inhibited. On head tilt to the opposite side, the converse occurs. As a result, the eyes rotate around the visual axis in a direction opposite the head tilt. ^1–16 However, this compensatory eye movement, ocular counterrolling (OCR), compensates for only approximately 10% to 20% of head tilt. ^{1–9,11,14–16}  

The Bielschowsky head tilt phenomenon (BHP) is used to diagnose superior oblique (SO) palsy and is characterized by an elevation of the paretic eye when the head is tilted toward the shoulder on the side of the lesion. The hypertropia is greater with head tilt to the ipsilesional than to the contralesional shoulder in patients with SO palsy, ^17,18 because of the loss of downward torque of the SO muscle. This phenomenon has traditionally been explained by a deficient contraction of the SO muscle during OCR; accordingly, BHP is considered the defining clinical criterion for SO palsy. ¹⁸ OCR is primarily mediated by changes in forces generated by the SO and inferior oblique (IO) muscles rather than by the vertical rectus muscles. ¹⁹ Clinically encountered cases with a huge BHP can be accounted for mainly by innervational changes of vertical rectus muscles, although the reason for these innervational changes have remained enigmatic. 

We previously reported in a small series of patients with SO palsy that for static OCR (s-OCR), the gain (defined as the mean percentage of compensatory s-OCR relative to the head tilt angle) in the paretic eye was less with ipsilesional than contralesional head tilt. ²⁰ Based on these preliminary data, we examined the relationship between BHP and s-OCR in patients with SOP, asking whether the BHP can be explained by the decreased s-OCR. We also analyzed the effect of IO-weakening surgery on s-OCR. 

Methods

Subjects

Thirty-three patients with unilateral decompensated or early-onset SO palsy were recruited in the study. The mean (SD) age was 40.1 (20.2) years, with a range of 9 to 72 years. All patients underwent a complete ophthalmic examination including refraction. Visual acuity was 20/20 or better. The diagnosis of decompensated SO palsy was based on clinical criteria of incomitance in the diagnostic gaze positions, including BHP, Hess red-green screen test, and characteristic clinical findings—that is, an unclear time of onset, gradual increase in diplopia, and photographs or a recollection of an abnormal head posture during childhood. The elevation of the paretic eye on ipsilesional head tilt, which is defined as BHP, was measured with the alternate prism and cover test, on head tilt 30° to the shoulder during fixation on a distant target (5 m). Patients with a clear cause of their vertical strabismus, such as stroke, myasthenia gravis, cerebellar degeneration, closed-head injuries, trochlear nerve palsy, or skew deviation as part of the ocular tilt reaction, were excluded. As a control, we recruited 11 normal subjects with a mean (SD) age of 30 (15) years. The study was conducted in accordance with the Declaration of Helsinki. 

Measurement System

A mouthpiece was made of dental impression compound type 1 (Kerr, Orange, CA) for each subject, to fix the measurement device to the head. A charge-coupled device (CCD) video camera (Model ET-110 [½ inch, 768 × 494 pixels]; Newopt, Kawasaki, Japan), was placed in front of the eye using two camera-mounting brackets. The iris was evenly illuminated by six infrared light-emitting diodes (LEDs). Cross-polarized filters were placed in front of the LEDs and the camera lens to minimize unnecessary reflection from the cornea. To create a target for fixation, a miniature laser pointer (Tan-chung; Steel, Yang-mei, Taiwan) was fixed parallel to the axis of the CCD camera. The total weight of this device was 314 g. 

A hologram filter, emitting laser light and producing four dots surrounding a central dot on the wall, was placed in front of the laser pointer. Through an infrared-reflecting mirror positioned in front of the camera, the subjects were requested to fix on the central dot. This enabled the examiner to control the subject's head position, not only in the roll plane but also in the horizontal and vertical planes. ¹⁶ At first, we adjusted the head position of the subjects horizontally and vertically so that the central dot coincided with the center of a protractor put on the wall at each subject's eye level. Then, using one of the four surrounding dots and a protractor, we adjusted the head position to the required angle (Fig. 1). 

Figure 1.

 

The head-mounted measurement device. The device was fixed to the subject's head with a mouthpiece (A). A charge-coupled device (CCD) video camera (B) was placed in front of the subject's eye and held stationery with two fixing brackets (C). The iris of the subject was evenly illuminated by six infrared light-emitting diodes (D) coaxial to the CCD camera. Through a hot mirror (E), the subject fixed his or her gaze on a target that was projected onto a wall with a laser pointer (F).

Measurement Procedure

While the subjects bit the mouthpiece firmly, we moved the head in the yaw plane 0°, 10°, 20°, 30°, and 40° to the shoulder, with occlusion of the other eye. To eliminate any confounding effect of torsional angular acceleration during the OCR, or the aftereffect of dynamic OCR, the head tilt angle was chosen in random order. Approximately 20 seconds after reaching the required head tilt angle, ¹⁶ digital images of the eye were captured with a video board (Snappy; Play Inc., Cordova, CA), as an 8-bit grayscale image. The video images were also recorded with a DVD recorder (digital versatile disc recordable for off-line analysis). The video images were converted into bit-map images (640 × 480 pixels; InterVideo WinDVD 4; Corel Corp., Ottawa, ONT, Canada) and stored on a personal computer. One session consisted of measurements for head tilt angles of 0°, 10°, 20°, 30°, and 40° randomly ²¹ and sessions were performed three times. Measurements were performed in a dimly lit room to minimize the influence of ambient illumination on the images of the pupil and visually driven torsional eye movements. Refractive errors of the subjects were uncorrected, and no miotic drugs were used to control the size of the pupil. 

Image-Processing Procedure

Threshold levels were determined to extract a characteristic iris pattern and corneal light reflex (Image; Scion Corp., Frederick, MD), and then the X–Y centroid coordinates of images that were selected for clarity were calculated (Fig. 2). Using this calculation, a subpixel resolution was possible. The threshold level was not altered throughout the analysis. Torsional eye position was measured using the changes in the angle of a line connecting the two centroids of the iris pattern and the corneal light reflex. 

Figure 2.

 

Scheme of video-based analysis. On the captured digital images (A), threshold levels that indicate the corneal light reflex (B, arrow) and the characteristic iris pattern (C, arrow) were determined. The X–Y coordinates of the centroids of the images were calculated. The torsional eye movement was evaluated by measuring the change in the inclination of the line connecting the two centroids on the images.

Calculation of s-OCR

Because there is a sinusoidal relationship between the amplitude of s-OCR and the inclination angle during whole body rotation (0°–360°) in the roll plane, ^14,22,23 we fitted the data to equation 1 (Fig. 3). Then, the angle of head tilt giving the maximum s-OCR was 90 (270)°.  

 

where θ, a, and b represent the head-tilt angle in degrees, amplitude of s-OCR and offset, respectively. On average, the obtained mean (SD) amplitude of s-OCR (a) was 10.9 (2.6°) with a range of 5.3° to 15.6° in normal control subjects. The offset (b) was 0.6°. 

Figure 3.

 

Relationship between inclination of the head tilt angle and amplitude of s-OCR in control subjects. Data from three measurement sessions in 11 normal subjects are shown. The fitted sin curve is f (θ) = 10.9 · sin (θ · π/180°) + 0.6.

IO-Weakening Surgery

s-OCR was measured in eight patients with SO palsy before and after IO-weakening surgery. The IO muscle was disinserted, tied with the double-armed locking bite technique with 6-0 Vicryl (Ethicon, Sommerville, NJ), and then fixed to the sclera 4 to 5 mm posterior to the temporal border of the insertion of the inferior rectus (IR) muscle, regardless of the degree of hyperdeviation in the primary position. 

Statistical Analysis

Results of the statistical analyses with a level of P < 0.05 were considered significant (JMP software; SAS, Cary, NC). 

Results

s-OCR in Response to Ipsi- and Contralesional Head Tilt

The mean (SD) amplitude of s-OCR on ipsi- and contralesional head tilt was 6.3 (3.5)° and 11.3 (3.9)°, respectively, in patients, and in control subjects, 10.9 (2.6)° (Fig. 4A). The mean amplitude of s-OCR for paretic eyes on ipsilesional head tilt was significantly decreased compared with contralesional head tilt (P < 0.001) and with control subjects (P < 0.001). The mean (SD) amplitude of s-OCR of the fellow eye on ipsilateral head tilt did not significantly differ from that on contralateral head tilt: 12.2 (4.4)° for ipsilateral and 11.0 (4.0)° for contralateral head tilt, P = 0.24] nor from the control subjects (P = 0.16; Fig. 4B). 

Figure 4.

 

Amplitude of s-OCR of the paretic (A) and opposite eye (B) on head tilt. No significant difference in the amplitude of s-OCR of the paretic eye was found between control and contralesional head tilt, but a significant difference was noted between ipsi- and contralesional head tilt. No significant difference in the amplitude of s-OCR of the opposite eye was found between control and contralateral head tilt and between ipsi- and contralateral head tilt.

Relationship between s-OCR and BHP

There was no significant correlation between the hyperdeviation on ipsilesional head tilt and the amplitude of s-OCR for paretic eyes (r ² = 0.04, P = 0.29; Fig. 5). However, there was a significant relationship (r ² = 0.19, P < 0.001) between the difference in hyperdeviation with ipsilesional head tilt and in the head-upright position against the amplitude of s-OCR in paretic eyes on ipsilesional head tilt (slope of 0.69; Fig. 6). 

Figure 5.

 

Relationship between the amplitude of s-OCR during ipsilesional head tilt of the paretic eye and hyperdeviation of the paretic eye during ipsilesional head tilt (BHP). The line indicates linear regression through all data points, with the slope not significantly differing from 0 (P = 0.29, r ² = 0.04, correlation coefficient: −0.19, y = 18.4 − 0.31x).

Figure 6.

 

Relationship between the difference in hyperdeviation measured with ipsilesional head tilt and in an upright position against the amplitude of s-OCR in paretic eyes during ipsilesional head tilt. The line indicates linear regression through all data points, with the slope significantly differing from 0 (P = 0.01, r ² = 0.19; correlation coefficient: −0.43, y = 9.9 − 0.69x).

Effect of IO-Weakening Surgery on s-OCR of the Paretic Eye on Ipsi- and Contralesional Head Tilt

IO-weakening surgery of the paretic eye significantly decreased not only the mean hyperdeviation in the upright head position (13.4 [5.7]° to 8.1 [6.3]°, P = 0.003; Fig. 7A) but also the mean hyperdeviation with ipsilesional head tilt (17.7 [5.0]° to 11.1 [5.7]°, P < 0.001; Fig. 7B). The difference in hyperdeviation between ipsilesional head tilt and head upright was not significantly decreased (4.3 [3.6]° to 2.9 [2.2]°, P = 0.31; Fig. 7C). IO weakening resulted in a significant decrease in the mean amplitude of s-OCR with contralesional head tilt (11.3 [3.2]° to 6.9 [3.3]°, P < 0.01; Fig. 8B) but did not change s-OCR with ipsilesional head tilt (6.9 [3.6]°to 6.1 [2.5]°, P = 0.69; Fig. 8A). The mean change in the amplitude of sOCR was 4.4 (2.8)° for contralateral and 0.7 (4.3)° for ipsilateral head tilt. 

Figure 7.

 

Change in hyperdeviation in upright (A) and head-tilt (B) positions after IO weakening surgery. A significant decrease in hyperdeviation of the paretic eye was found in the upright position (P < 0.0001) and also with ipsilesional head tilt (P = 0.003). The Difference in hyperdeviation between head-tilt and head-upright positions (C) did decrease significantly (P = 0.31).

Figure 8.

 

Change in amplitude of s-OCR in ipsi- (A) and contralesional (B) positions after IO-weakening surgery. No significant decrease in the amplitude of s-OCR of the paretic eye was found when the head was tilted ipsilesionally (P = 0.69), but it was significant during contralesional head tilt (P = 0.003).

Discussion

The main findings of this study are that the amplitude of s-OCR was significantly decreased in paretic eyes during ipsilesional head tilt, but normal during contralesional head tilt. Furthermore, after IO weakening surgery, there was a significant decrease in s-OCR for contralesional head tilt. Our previous small case series involving SO palsy showed that the s-OCR gain (i.e., the ratio of the degree of s-OCR to the inclination of the head tilt angle in paretic eyes on ipsilesional head tilt with SO muscle atrophy) was significantly decreased in comparison with that in patients without SO muscle atrophy. ²⁰ Thus, in accordance with several previous studies, ^1–12,14,16 we inferred that the oblique muscles play an important role in OCR during head tilt in the roll plane and concluded that the amount of s-OCR is a useful index for assessing the function of the SO muscle. 

There have been few studies on the measurement of s-OCR in patients with SO palsy. Simonsz et al. ¹¹ reported no significant relation between hypertropia in paretic eyes on ipsilesional head tilt and the amplitude of s-OCR. They described little difference in the amount of s-OCR of the paretic eye between ipsi- and contralesional head tilt in 23 patients with clinically diagnosed SO palsy and concluded that if an SO is paretic, there is less s-OCR. They recorded eye movements under a binocular viewing measurement condition using photographic methods and found 4.6 (2.1)° for ipsilesional head tilt and 4.3 (2.0)° for contralesional head tilt. In their experiment, 7 (30%) patients showed a decreased amount of s-OCR, whereas 7 (30%) revealed an increased s-OCR during ipsilesional head tilt. A possible reason for the difference between their and our results is the difference in measurement conditions. We measured under the monocular viewing condition, in which case s-OCR is less likely to be influenced by any cyclovergence movements that occur during binocular viewing. As Misslisch et al. ²⁴ have pointed out, with convergence, OCR is suppressed, presumably to help to maintain stereoscopic vision. Another reason may be a difference in the cause of SOP between the two groups (Simonsz et al. ¹¹ included acquired SOP) since there was good agreement of the maximum and mean amplitude of s-OCR in normal subjects between both studies. 

BHP is a clinical test for determining the side affected by the SO palsy. Although the biomechanical basis for BHP is not fully understood, it is usually attributed to a loss of downward torque of the palsied SO. Figure 6 shows that as the amplitude of s-OCR decreased, the amount of difference in hyperdeviation between head-tilt and head-upright increased. This occurred because there was more superior rectus (SR) activation needed to counteract the lack of s-OCR. In our study, however, the degree of hyperdeviation did not correlate with the amplitude of s-OCR on ipsilesional head tilt, raising questions about the precise relationship between BHP and the function of the SO muscle. Kono et al. ²⁵ also concluded that the cross-sectional area of the SO muscle on magnetic resonance imaging (MRI) does not account for the variation in BHP in patients with SO palsy when the diagnosis is based on clinical features alone. Kushner ²⁶ claimed that the BHP may be positive in patients with vertical strabismus due to dissociated hyperdeviation, previous vertical muscle surgery, skew deviation, myasthenia gravis, or small nonparalytic and hyperdeviations associated with horizontal strabismus. Consequently, a positive result in a three-step test is not specific for SOP. Recent studies have shown that SO palsy diagnosed on clinical criteria alone is a heterogenous group of diseases. SO palsy may not necessarily be neurologic, because abnormalities of the tendon of the SO muscle, ^5–7,8,27 of the orbital pulleys, ^28,29 or other mechanical causes may induce vertical strabismus. Patients in whom MRI was used to diagnose SO palsy based on clinical features alone commonly showed an SO muscle with a normal cross-sectional area and normal contractility. In other words, these patients have a pattern of cyclohyperdeviation that can mimic SO palsy. ^28,29 Incomitant vertical strabismus can mimic the pattern of SO palsy. Simulations with a biomechanical simulator (Orbit 1.8; Eidactics, San Francisco, CA) suggest that vertical mislocation of the horizontal rectus muscle pulley can produce a clinical pattern of SO underaction and IO overaction without postulating any SO weakness. ³⁰ Kono and Demer ²⁹ confirmed and extended this finding in patients with incomitant vertical strabismus diagnosed as SO palsy, but demonstrated by MRI as rectus pulley heterotopy. 

Biomechanical simulations predict that the SO muscle itself generates too small a vertical force to account for the marked hypertropia on ipsilesional head tilt typically observed in SO palsy. Robinson ³¹ showed that a loss of downward torque of the SO muscle would only explain a BHP of 2.8°, whereas most patients had much larger values. ³² Quaia et al. ³³ based on mechanical simulations, suggested that variations in the anatomic location of the insertion of the SO tendon can have a large impact on the pattern of deviation in SOP. Other secondary innervational or mechanical changes of the palsied muscle and the vertical rectus muscle are also suggested as causes of intersubject differences in BHP. Central adaptive mechanisms could amplify the otolith reflex to reduce the compensatory head tilt required for binocular single vision. This mechanism would reflect increased activation of the SR of the contralesional eye and of the IR of the ipsilesional eye, thus decreasing the hyperdeviation. The action of the vertical rectus muscles would then dominate that of the IO muscle, enabling binocular single vision with a relatively small head tilt. This mechanism could potentially lead to hypertropia on ipsilesional head tilt. ^34,35 However, in our study the preoperative mean amplitude of s-OCR in the paretic eye on contralesional head-tilt was not significantly different from the normal control. In addition, no significant change in the mean amplitude of s-OCR in the paretic eye on ipsilesional head-tilt was noted, although hyperdeviation on head tilt significantly decreased after surgery. Based on our study, such an adaptation mechanism ³² does not account for the large hyperdeviation encountered clinically, but innervational or mechanical alternations may affect the SR of the paretic eye on ipsilesional head tilt. 

Why was there a significant decrease in hyperdeviation on ipsilesional head tilt after IO-weakening surgery, even though the amplitude of s-OCR on ipsilesional head tilt did not change significantly? Recently, Kushner ³⁶ presented an interesting hypothesis of anticompensatory torsional movements that eliminates dynamic compensatory counterrolling and occurs in the direction of head tilt. ^{5,6,8–10,12,13,15} Shan et al. ³⁷ also show that torsional optokinetic nystagmus (OKN) in monkey include quick phases which is the equivalent of anticompensatory torsional movements. According to Kushner's hypothesis, during the initial anticompensatory torsional movement, the IR and IO of the paretic eye are stimulated on ipsilesional head-tilt; thereafter, extorsional movements occur due to the relaxation of ipsilesional SO. Considering this hypothesis and our findings, a postoperative decrease of hypertropia on ipsilesional head tilt could be interpreted as an overpowering of the IR of the paretic eye, because the IR of the paretic eye is unopposed during anticompensatory torsional movement after IO-weakening surgery. 

IO-weakening resulted in a greater loss of contralesional s-OCR than ipsilesional. This finding indicates that IO muscle largely contributes as an extorter during torsion with contralateral head tilt, and that OCR is primarily mediated by change in forces generated by the SO and IO muscles rather than by the vertical rectus muscles. ¹⁹  

We found a significant negative correlation between the amplitude of s-OCR on ipsilesional head tilt of the paretic eye and the difference in hyperdeviation between ipsilesional head tilt and head-upright position. This result demonstrated that the amplitude of s-OCR is primarily reflected in the difference in hyperdeviation between ipsilesional head tilt and a head-upright position, not on ipsilesional head tilt alone. The amplitude of s-OCR in six patients (86%) with a 10° or larger difference in hyperdeviation between head tilt and head-upright position was smaller than the lower limit of the 95% CI of s-OCR amplitude. If the s-OCR amplitude in ipsilesional head tilt position represents SO muscle function, the difference in hyperdeviation between ipsilesional head tilt and head-upright position may be an alternative indicator of SO function. 

There are some limitations to our study. We included patients with decompensated SO palsy, but not patients with acute SO palsy. Therefore, the associated contracture of agonist and antagonist muscles could have affected the s-OCR measurements. This effect could have resulted in a nonsignificant relation between the hyperdeviation on ipsilesional head tilt and the amplitude of s-OCR, because the amount of hyperdeviation can be accounted for mainly by innervational changes of the vertical rectus muscles. 

In summary, we found no relationship between s-OCR and hypertropia on ipsilesional head tilt, which suggests that hypertropia on ipsilesional head tilt does not specifically reflect SO function in clinically diagnosed SO palsy. The difference in hypertropia between the head tilt and the upright position, however, may be a better indicator of SO function. 

Footnotes

 Supported in part by Grant-in-Aid 19592023 from the Ministry of Education, Science, Sports, Culture, and Technology of Japan; and the Koyama Fund.

Footnotes

 Disclosure: I. Hamasaki, None; S. Hasebe, None; T. Furuse, None; H. Ohtsuki, None

The authors thank David Zee, Howard Ying, Xiaoyan Shan, and Jing Tian for help with the manuscript.