Prevalence of visual stress in dyslexia and controls (1) - as good as it gets

Kruk R, Sumbler K, Willows D. Visual processing characteristics of children with Meares-Irlen syndrome. Ophthalmic Physiol Opt. 2008 Jan;28(1):35–46. pubmed link
This paper is by some distance the best in trying to control for all biases that are problematic in case control studies.  Helen Irlen has claimed that 'Irlen Syndrome' or visual stress is found in 12-14% of the general population and 46% of dyslexics. However, this figure does not come from the peer reviewed literature and can not be evaluated, or in my opinion be trusted.

In this series of posts I will be reviewing those studies that compare the prevalence of visual stress in individuals with dyslexia and matched controls - so called case control studies. Such studies a prone to are number of biases that can render the results totally unreliable. Please see the previous post for an example.
The important biases are

• Selection bias. If the cases and the controls are drawn from different populations the difference in the prevalence visual stress may be amplified. For example, cases with dyslexia drawn from a dyslexia club and controls drawn from a classroom setting. 
• Ascertainment bias. a) Assessor bias -if the experimenters are not blinded to the reading status of the subjects being tested they may be, consciously or subconsciously, more likely diagnose visual stress in the poor readers. b)Response bias- bias may also be on on the side of the subjects. For example the subjects may have already been made aware of the potential benefits of coloured lenses and overlays before the assessment; a particular problem when cases are drawn for specialist dyslexia clinics of dyslexia clubs.

In this study, cases (18) and controls (18) were drawn from four schools in Toronto.
Cases were individuals with a reading age two years less than their chronological age.
Controls were matched for age but were average of above average  readers.

Screeners were blinded to the reading status of the subjects.  The test used was the Irlen Reading Perceptual Scale (IRPS) which has two parts. Different screeners, who where unaware of the results of the first part, where used for the second part of the test.
Subjects were also tested for the visual processing subtype of dyslexia which includes indicators of visual discrimination, visual memory, visual-motor integration and pattern analysis.

Results

The results were surprising.

• 10/18 or 55% of controls had visual stress but 8/18 or 44.4% of controls Odds ratio=0.64  95%CI 0.17-3.38
• If questionable diagnoses are included the results become more even with 13 in each group or 72% having visual stress. Odds ratio=1.0 95%CI 0.23-4.3
• There was no association between visual stress as measured with the IRPS and the visual processing subtype of dyslexia. 
• Visual stress was not a predictor of reading or spelling difficulties.

The data is also shown in the forest plot below. It can be seen that because the sample sizes are small, the data could be consistent with visual stress being more prevalent in the control group or the cases with reading impairment. It certainly does not fit with Helen Irlen's assertion of 12-14% of the general population and 46% of dyslexics.

Extrapolating to a population basis, visual stress is at best a very weak risk factor. Assuming a prevalence of dyslexia of 5%. Out of a 1000 individuals in the population, we would expect 50 dyslexics of whom about 44% or 22 would have visual stress. Of the 950 'normal readers'  522 would have visual stress. So -there would be 24 normal readers with visual stress for every individual with dyslexia and visual stress.



Conclusion
This study is a major problem for proponents of the use of colour to treat visual stress in order to ameliorate reading difficulties.  The authors argue that it visual stress (or MISViS as they say) is a separate condition from dyslexia. However, if visual stress is a risk factor, whose treatment can ameliorate reading difficulties, we would expect it to be more prevalent in the population with dyslexia. This study does not support that view. Worse still, this is about as good as the evidence gets.