Draft proposal of Initial analyses of available data

Introduction and background

From the development of the bony orbit through to the complex interconnection of neuronal circuitry required for stereoscopic vision, the factors controlling growth and development of the visual system in humans are poorly understood. In such a sensitive and plastic system, developmental anomalies are extremely common. In many circumstances small but timely interventions can result in significant improvements in vision which serve that individual for life.

Strabismus (squint) describes misalignment of the visual axis of one eye with respect to the other. It is thought to affect between 2 and 4% of the population and in most cases results in loss of vision and/or loss of stereoscopic vision1. Some squints are convergent and some divergent. In some cases it is not clear why the eye is divergent rather than convergent or visa-versa. Convergence is closely related to many forms of convergent squint as is the refractive condition of the developing eye2. Similarly, the distance between the eyes in the horizontal plane is known to alter the amount of convergence required for binocular vision and the degree of stereopsis in the general population can vary greatly. Therefore, it is possible that by altering the requirement for convergence, the physical distance between the eyes in the horizontal plane is a risk factor for squint and for varying grades of stereoscopic vision.

The factors influencing the growth of the bony orbit and its relationship to development of the globe are also poorly understood. However, it is known that extreme anomalies of globe growth can result in significant growth changes in the developing orbit. For example, in infants with either absent (anophthalmos) or very small globes (microphthalmos) the orbit is known to under-develop such that tissue expanders are sometimes used to treat asymmetry. Conversely, for infants with marked enlargement of the globe (Buphthalmos) which is often related to congenital glaucoma, the bony orbit is enlarged3.

It is therefore evident that extreme variations in the growth of the globe can influence the growth of the bony orbit and resulting facial character. Perhaps more subtle variations in globe size also have an effect on the development of the bony orbit.

Proposed study

ALSPAC is an ongoing population-based birth cohort into which over 14,000 mothers-to-be were recruited in the Southwest of England (http://www.bristol.ac.uk/alspac/). Approximately 11-12,000 children and their families are still participating and the "children" are now on the verge of adulthood (17 - 19 years). ALSPAC has assembled detailed exposure and developmental data on these children and their families in addition to genetic information leading to the production of numerous peer reviewed publications in prestigious journals4.

Data relevant to this proposed study include 3D facial scans for 4200 individuals of the ALSPAC cohort when aged 15, processed and landmarked by Professor Richmond's group in Cardiff. The 3D facial scans were captured using 2 high resolution Konica/Minolta laser scanners. These scans provide exquisite detail of the surface characteristics of each participants face and allow measurement of facial features and comparison with age matched controls to an accuracy of 0.3mm. These data have been 'landmarked' over various periocular structures to allow measurements of anatomical structures such as inter-canthal distance and new landmarks can be created for each scan to measure an infinite number of facial characteristics.

Further relevant data includes assessments of ocular biometry, alignment and stereoscopic vision measurements for the same 4200 children with facial scans, collected at ages 7 - 15 years, thus providing a uniquely detailed dataset of ocular development and facial characteristics in a large sample of children.

We will investigate the hypothesis that mild to moderate variations in globe size alter the growth of periocular structures. From those ALSPAC children with 3D facial scans and ocular biometry for each eye collected at age 15 (n = 2837), approximately 5% (134) have an interocular difference in axial length of 0.5 mm or greater. We will obtain a Left vs Right asymmetry measurement (as in figure 1) for the periocular landmark data and compare the asymmetry of the two groups and therefore the degree of facial asymmetry with the degree of ocular size asymmetry.

Additionally, we will compare the same periocular landmarked data in children (obtained at age 15) with orthoptic measurements obtained at age 7 years (strabismus) or stereoscopic vision (obtained at 7, 11 and 12 years), to investigate the relationship between horizontal distance between the visual axes and the incidence of squint and reduced levels of stereopsis (depth perception). We anticipate that the majority of children with 3D face data will have been examined at age 7: of these approximately 3.5% (estimated n = 140) will have had a strabismus at that age and are highly likely still to be strabismic at 15 as the condition does not spontaneously resolve and even surgery rarely removes all misalignment. We will investigate the hypothesis that the occurrence of ocular misalignments (manifest squints) is modified by the distance between the ocular axes, for example that the risk of convergent strabismus increases as the horizontal distance between the eyes decreases and conversely that the risk of divergent strabismus increases as the eyes are displaced further apart. Stereopsis has been tested at 7, 11 and 12 years and we anticipate that the majority of children with 3D facial data at 15 will have had their stereopsis tested at 7 years, and again at 11 years. We will investigate the hypothesis that, in children with no strabismus, the accuracy of depth perception is related to the horizontal distance between the eyes.

We will perform a variety of statistical analyses looking for associations between these parameters and ocular landmark measurements. The details of the individual tests will be clarified when feasibility and numbers for each individual dataset are ascertained. We anticipate using univariate and subsequently multivariate analyses using appropriate covariates including sex etc. Dr Kate Northstone will act as "data buddy" as part of her funded programme of work with CW and will advise on analytic approaches as needed. Data transfer will take place using a specific procedure designed to collaborate with Prof Richminds group, who unusually have the alspac clinicID as the identifier on the face-shape data they acquired. Thus KN will send Prof Richmond the clinicIDs of all participants for whom we have the specified vision variables; Prof R will then send KN the derived face landmark variables; KN will attach to this file the vision data and will remove clinicID and replace it with a collaborator ID. This stand-alone file can then be analysed in Cardiff or Southampton as appropriate.

Additional statistical support will be sought where necessary from Dr Sarah Ennis from the Bioinformatic Unit at Southampton General Hospital, and Professor Clive Osmond, Professor of medical statistics at Southampton's MRC Epidemiology resource centre

We anticipate acquiring sufficient numbers to investigate the associations of various ocular parameters with periocular facial landmarks. This will provide some exiting information about the complex interaction between physical ocular anatomy, visual development and the interaction of orbital and ocular growth. This could provide some interesting epidemiological data in addition to a greater understanding of visual system development. It is anticipated that this work will lead to the publication of research papers in peer reviewed journals and may provide preliminary data for further work in this cohort perhaps using genetic data.