Markers of cognitive function in individuals with metabolic disease: Morquio syndrome and tyrosinemia type III

ABSTRACT

We characterized cognitive function in two metabolic diseases. MPS–IVa (mucopolysaccharidosis IVa, Morquio) and tyrosinemia type III individuals were assessed using tasks of attention, language and oculomotor function. MPS–IVa individuals were slower in visual search, but the display size effects were normal, and slowing was not due to long reaction times (ruling out slow item processing or distraction). Maintaining gaze in an oculomotor task was difficult. Results implicated sustained attention and task initiation or response processing. Shifting attention, accumulating evidence and selecting targets were unaffected. Visual search was also slowed in tyrosinemia type III, and patterns in visual search and fixation tasks pointed to sustained attention impairments, although there were differences from MPS–IVa. Language was impaired in tyrosinemia type III but not MPS–IVa. Metabolic diseases produced selective cognitive effects. Our results, incorporating new methods for developmental data and model selection, illustrate how cognitive data can contribute to understanding function in biochemical brain systems.

KEYWORDS:

• Attention
• developmental disorder
• language
• morquio
• tyrosinemia

Acknowledgements

The authors would like to thank all of the individuals with metabolic disease and their families who supported this research, the many schoolchildren who participated in the control sample, and the teachers and school administrators who made this possible.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Model selection using AIC is different from that using p-values, but not difficult to understand. AIC is preferred for model comparison because, unlike p-values, AIC balances fit and the number of model parameters when choosing models. In brief, better models produce smaller AIC values, but the absolute AIC values are not interpretable. Instead, the change in AIC (ΔAIC) between models is meaningful and captures the weight of evidence for each model (rather than being subject to a cut-off, like p-values). Evidence for a model starts to be clear if the ΔAIC exceeds 2. If ΔAIC between the “best” model and alternative models is less than 2 then the two models are substantially equivalent. When ΔAIC is between 2 and 10 there is decreasing support for an alternative model. A model with a ΔAIC >10 has essentially no support. For models where the ΔAIC is less than 2, it is reasonable to favour the least complex model (i.e. model with fewest parameters/variables). Favoured models contain terms that are important in accounting for data. This is parallel to significant effects in an analysis using hypothesis testing. For example, if a highly rated model has a term for group but no interaction, this is parallel to a significant main effect of group and a non-significant interaction.

Comparisons can be assisted by calculating Akaike weights (AICw; Burnham & Anderson, 2002). AICw expresses the relative probability that a model is the best in a particular set, considering only the models from that set. It measures the weight of evidence for the models being compared. When values are relatively equal across two or more models, they are all relatively good models of the data. If one model has a high value, and the others are low, there is a model that is clearly better.

2 Specifically, for values above the mean, and , where = upper boundary of the smoothed prediction interval; control mean = the mean predicted by smoothing control mean values; and = individual patient’s mean value. Values below the mean were calculated in the same way except the 95% prediction interval boundary used the lower boundary since boundaries were not necessarily symmetric.

3 This is not to deny that lexical and sub-lexical mechanisms can display dissociations also during development (see summary in Castles, Bates, & Coltheart, 2006), but these dissociations do not exclude the possibility that there are also learning dependencies between components that could be revealed by developmental studies in a way that is not possible in adults (as Castles et al., 2006, p. 881, note, and as the methods of Pitchford & Funnell, 1999, illustrate).

Additional information

Funding

This work was supported by Economic and Social Research Council (ESRC) CASE studentship to James Blundell in collaboration with the Society for Mucopolysaccharide Diseases (MPS Society); and by Birmingham Children’s Hospital Charities [grant number BCHRF 230].