**Purpose**:
The aim of this study was to explore the relationship between compliance with preoperative posturing advice and progression of macula-on retinal detachment (RD) and to evaluate whether head positioning or head motility contributes most to RD progression.

**Methods**:
Sixteen patients with macula-on RD were enrolled, admitted to the ward, and instructed to posture preoperatively. The primary outcome parameter was compliance, which was defined as the average head orientation deviation from advised positioning. Secondary outcome parameters included the average rotational and linear head acceleration. The head orientation and acceleration were measured with a head-mounted inertial measurement unit (IMU). Optical coherence tomography (OCT) imaging was performed at baseline and during natural interruptions of posturing for meals and toilet visits to measure RD progression toward the fovea.

**Results**:
The Spearman correlation coefficient with RD progression was 0.37 (*P* = 0.001, *r*_{s}^{2} = 0.13) for compliance, 0.52 (*P* < 0.001, *r*_{s}^{2} = 0.27) for rotational acceleration, and 0.49 (*P* < 0.001, *r*_{s}^{2} = 0.24) for linear acceleration. The correlation coefficient between RD progression and rotational acceleration was statistically significantly higher than the correlation coefficient between RD progression and compliance (*P* = 0.034).

**Conclusion**:
The strength of the correlation between RD progression and compliance was moderate. However, the correlation between RD progression and rotational and linear acceleration was much stronger. Preoperative posturing is effective by reducing head movements rather than enforcing head positioning.

**Translational Relevance**:
Monitoring the efficacy of preoperative posturing in macula-on RD using OCT and IMU measurements shows that a new and combined application of these technologies leads to clinically relevant insights.

^{1,2}Visual acuity may be severely affected if the RD extends to the macula.

^{3–5}To prevent macular involvement, preoperative posturing is prescribed while patients are waiting for surgery. Patients with macula-on RD are prescribed bed rest to reduce head and eye movements and related fluid currents.

^{6–13}Additionally, patients are positioned supine when RD is located in the superior quadrants of the retina and upright for RD in the inferior quadrants to address the effect of gravity. To improve the compliance with this posturing advice, in some clinics patients are hospitalized during the preoperative period. An alternative approach is to provide surgery on a 24-hour, 7-days-per-week basis. As both approaches are expensive policies, the understanding of the effectiveness of preoperative posturing warrants further study. Recently, we used optical coherence tomography (OCT) to demonstrate that preoperative posturing reduces the progression of macula-on RD by comparing posturing with interruptions for meals and other short breaks.

^{14}However, the strength of the relationship between compliance to preoperative posturing and RD progression is as yet unknown.

^{15–17}In this study, we used such sensors to measure the head orientation as well as the head's rotational and linear motility in patients with macula-on RD. Because the density differences between the retina and subretinal fluid are rather small, we would expect gravity to play a limited role in the progress of RD.

^{18}Therefore, we hypothesized that head movements and eye movements contribute more to progression of RD than does head positioning.

^{14}The recordings of head orientation and head motility were performed only in the patients enrolled in the small additional cohort, which is presented in this report. All patients were hospitalized and examined in the Rotterdam Eye Hospital, Rotterdam, The Netherlands, and all provided written informed consent. The study was conducted in accordance with the tenets of the Declaration of Helsinki.

^{14}In brief, patients diagnosed with macula-on RD were admitted to the ward for posturing while they were waiting for surgery the same day, the next day, or occasionally the day after. Posturing consisted of two parts: bed rest and positioning. Patients with RD located mainly in the superior quadrant were positioned supine; patients with RD in the temporal quadrant on the temporal side of the affected eye, patients with RD in the nasal quadrant on the nasal side, and patients with RD in the inferior quadrant were instructed to sit upright. Patients were allowed to interrupt their posturing for meals, toilet visits, refreshment in the morning, and surgeon's examinations. Such intervals offer an excellent opportunity to acquire prospective and comparative data in an ethically acceptable manner.

^{19}

^{14}The 95% limits of agreement of the intrarater variability of these distance measurements was ±58 μm.

^{14}The distance measurements on subsequent OCT scans were then used to calculate the RD border displacement and the average RD border displacement velocity (change in distance per hour) during posturing and interruption intervals. The latter measure adjusts for differences in interval duration and thereby enables a more consistent comparison between OCT measurements and average head orientation and head movements per measured interval. The average progression velocity from baseline was determined at each time point as well.

*X*-axis points toward the north, the

*Y*-axis toward the west, and the

*Z*-axis up, perpendicular to the earth's surface. To prevent gimbal lock, quaternions were used instead of Euler angles to describe the three-dimensional rotations. The quaternions were calculated using the Shimmer Matlab Instrument Driver software (Shimmer Sensing), which estimates orientation data using magnetic angular rate and gravity (MARG) filtering. MARG filtering is reported to achieve orientation accuracy levels with less than 0.8° static error and less than 1.7° dynamic error.

^{19}

^{2}at increments of 250°/s

^{2}and the linear acceleration thresholds between 0.25 and 10 m/s

^{2}at increments of 0.25 m/s

^{2}.

*U*test) was used to compare RD progression and IMU parameters between posturing and interruption intervals. We expected a monotonic but possibly nonlinear relationship between RD progression and the IMU parameters. Therefore, Spearman's correlation coefficient was calculated to describe the relationship between RD progression and average IMU parameters. The correlation analysis was performed for all measured intervals as well as for the progression from baseline. For all IMU parameters, a positive correlation demonstrates an association with RD regression and a negative correlation an association with RD progression. Statistical significant differences between correlation coefficients were tested according to the methods of Meng et al.

^{20}To determine whether the duration of follow-up (defined as the time between baseline OCT and last OCT measurement) influences the rate of RD progression from baseline, we also used the Spearman's correlation coefficient to describe the relationship.

*P*= 0.003). The median RD border displacement velocity during posturing intervals was −1 μm/h (IQR: −9 to 34 μm/h) and during interruptions −202 μm/h (IQR: −491 to 0 μm/h), which was statistically significantly different as well (

*P*< 0.001).

^{2}for rotational acceleration, and 0.8 m/s

^{2}for linear acceleration. The difference between posturing intervals and interruptions was statistically significant for all four IMU parameters (

*P*< 0.001).

*r*

_{s}) between RD progression and the IMU parameters were found for rotational acceleration (

*r*

_{s}= 0.52) and linear acceleration (

*r*

_{s}= 0.49). The

*r*

_{s}

^{2}can be interpreted as a proportion of explained variance if the IMU parameters and the RD progression are presented as ranked variables. The higher this proportion, the more variance is explained by a specific variable. The

*r*

_{s}

^{2}was 0.13 for orientation deviation from the advised positioning, 0.13 for the orientation deviation from optimal positioning, 0.27 for the rotational acceleration, and 0.24 for the linear acceleration. This means that rotational acceleration as well as linear acceleration seems to explain twice as much of the variance of RD progression than orientation deviation from advised or optimal positioning. The correlation coefficient between RD progression and rotational acceleration was statistically significantly higher than the correlation coefficient between RD progression and compliance (

*P*= 0.034; see also Table 3).

*r*

_{s}

^{2}= 0.27) for orientation deviation from optimal positioning, 0.68 (

*r*

_{s}

^{2}= 0.46) for rotational acceleration, and 0.72 (

*r*

_{s}

^{2}= 0.49) for linear acceleration. This means that the secondary IMU parameters are codependent with the primary IMU parameter (orientation deviation from advised positioning), but they are not the same.

*r*

_{s}= −0.36;

*P*= 0.001).

*P*= 0.76).

^{2}for rotational acceleration and 0.51 at a threshold level of 1.25 m/s

^{2}for linear acceleration (for full analysis, see Supplementary File S2). The increase of threshold levels did not result in substantially higher correlation coefficients than the correlation between RD progression and average IMU parameters per interval as presented in Table 3.

^{6–11}Apparently, a reduction of head movements is also beneficial to prevent RD progression.

^{21,22}The rotational velocity and acceleration of saccades are typically faster than those of active head rotations.

^{23–27}However, the radius of the head is greater than the radius of the eye, whereas the magnitude of saccades is smaller than that of head movements.

^{28}Therefore, the tangential linear acceleration of the components of RD may be in the same range. During a head rotation, the movement of the eye approximates a translational movement. The direction of acceleration and deceleration forces of the fluids at opposite sides of the eye will be almost parallel during rotational head movements and precisely parallel during linear head movements. As a result, the effect on fluid currents within the eye may be limited. During a saccadic eye rotation, however, the direction of acceleration and deceleration forces will be opposite on the opposite sides of the eye, which is likely to create strong fluid currents of both liquefied vitreous and subretinal fluid. Nevertheless, the number and strength of saccades can partly be predicted by the number and strength of head movements as measured in the current study.

^{29,30}Therefore, if saccades were to be measured independently from head movements, we expect that only a small additional part of the variance of RD progression could be explained.

^{31,32}We previously demonstrated that a small RD in the periphery has a higher progression risk, suggesting a difference in retinal adhesion as well.

^{14}Secondly, the amount of subretinal fluid and the shape of the detachment differs between RDs. It is expected that the retina reattaches faster in a flat RD than in a bullous RD with the same area of detachment because the subretinal fluid volume is smaller and will be reabsorbed earlier.

^{12,33–36}Thirdly, the size, number, and type of retinal breaks differ between RD patients, where a large horseshoe-shaped retinal tear is more likely to facilitate inflow of liquefied vitreous into the subretinal space than do small round holes.

^{12,13,37}Finally, the contractile properties of the detached, incompletely detached, or not detached vitreous differs among patients, mostly due to the effects of aging of the vitreous.

^{38,39}Progressive traction of contractile vitreous may detach the retina surrounding the retinal break, allowing more liquefied vitreous to enter the subretinal space.

^{12,13}Because of all these factors, head and eye movements can be only partly accountable for the variance in RD progression.

^{18}

^{15–17}after RD surgery when intraocular gas is used, after pneumatic displacement of submacular hemorrhages, and after corneal transplantation when air bubbles are used to facilitate attachment of the graft.

**J.H. de Jong**, None;

**K. de Koning**, None;

**T. den Ouden**, None;

**J.C. van Meurs**, None;

**K.A. Vermeer**, None

*. 2013; 120: 616–622.*

*Ophthalmology**. 1982; 100: 289–292.*

*Arch Ophthalmol**. 2006; 113: 1734–1742.*

*Ophthalmology**. https://doi.org/10.1136/bjophthalmol-2016-309097.*

*Br J Ophthalmol**. 2016; 41: 543–550.*

*Curr Eye Res**. 1973; 90: 117–120.*

*Arch Ophthalmol**. 1977; 18: 286–291.*

*Mod Probl Ophthalmol**. 1977; 55: 303–316.*

*Acta Ophthalmol (Copenh)**. 1980; 64: 94–97.*

*Br J Ophthalmol**. 2001; 24: 799–804.*

*Chang Gung Med J**. 2004; 24: 246–253.*

*Retina**. 1984; 98: 681–693.*

*Am J Ophthalmol**. 2014; 51: 15–31.*

*Ophthalmic Res**. https://doi.org/10.1016/j.ophtha.2017.04.004.*

*Ophthalmology**. 2014; 252: 739–744.*

*Graefes Arch Clin Exp Ophthalmol**. 2016; 36: 1081–1086.*

*Retina**. 2017; 124: 1510–1522.*

*Retina**. 2009; 86: 1187–1195.*

*Optom Vis Sci**. 2011; 2011: 5975346.*

*IEEE Int Conf Rehabil Robot**. 1992; 111: 172–175.*

*Psychol Bull**. 1984; 98: 681–693.*

*Am J Ophthalmol**. 2011; 10: 191–202.*

*Biomech Model Mechanobiol**. 1996; 6: 295–313.*

*J Vestib Res**. 1993; 86: 69–74.*

*Electroencephalogr Clin Neurophysiol**. 1989; 15: 529–543.*

*J Exp Psychol Hum Percept Perform**. Blacksburg, VA: Virginia Polytechnic Institute and State University; 2005. Thesis.*

*Linear and Angular Head Accelerations in Daily Life**. 2004; 48: 1973–1977.*

*Proc Hum Factors Ergon Soc Annu Meet**. 1975; 14: 468–469.*

*Invest Ophthalmol**. 2000; 131: 22–32.*

*Exp Brain Res**. 2001; 139: 266–277.*

*Exp Brain Res**. Los Angeles, CA: Elsevier Saunders; 2013: 2010–2018.*

*Retina*. 5th ed*. 1988; 95: 1385–1388.*

*Ophthalmology**. 2004; 49: 96–108.*

*Surv Ophthalmol**. 2013; 33: 105–110.*

*Retina**. 1996; 103: 1313–1324.*

*Ophthalmology**. Los Angeles, CA: Elsevier Saunders; 2013; 1: 447–464.*

*Retina**. 2011; 52: 5437–5440.*

*Invest Ophthalmol Vis Sci**. 2015; 17: 85-015-9988-z.*

*Biomed Microdevices**. 1987; 225: 89–93.*

*Graefes Arch Clin Exp Ophthalmol*