Commentary

Of interest to the clinician is an answer to the following question: do ethnic minority patients such as people of Caribbean origin have a different course of illness than their white British peers? This question clearly has face validity because ethnic minority status is associated with differences in pathway to care, service uptake, risk of misdiagnosis, treatment adherence, and cultural influences on perceptions of illness aetiology and treatment. The combined effect of these differences may well influence course and outcome. How helpful are the studies by Goater et al and Harrison et al?

Authors of both studies refer to previous publications for a clear description of the ethnic composition of their samples. Goater et al combine a group of patients of Caribbean and African origin on the basis of their appearance into a “black” group, whereas a group of Indian, Pakistani, and other Asians are grouped together on the basis on their place of origin into an “Asian” group. Finally, a group of patients of Greek, Turkish, Irish, British, and other European origin is again grouped on the basis of appearance into a “white” group. Of the white group, actually fewer than half were of British white origin. The question arises of how meaningful is the comparison between the black and white groups in this study. Both groups are defined on the basis of appearance, making the comparison more a racial than an ethnic one. The fact that more than half of the white group were of non-British origin means that differences due to ethnic minority status, if they existed, would be difficult to detect in this small sample.

Harrison et al were able to define a group of patients of Caribbean origin. Illness course in this group, however, was compared with all other ethnic groups lumped together. Although their previous publications suggest that the number of patients of African and Asian origin was small, no data are given on the origin of the white patients. Their groups, however, are clearly more homogeneous ethnically than the ones in the study of Goater et al and as such constitute a more relevant contrast clinically.

Neither study reports how large a difference in course they expected to find, nor how large their study samples needed to be to be able to detect such a difference. If we assume a clinically relevant effect size of 2—that is, the odds of a single episode/good recovery is 2 times higher in African-Caribbean patients—then the study by Harrison et al would have had approximately 40% chance of detecting it. On the other hand, they would have had 85% chance of detecting an effect of 3.

So what can clinicians conclude from these 2 studies? If we wish to know whether ethnic minority status, for example African-Caribbean ethnic group, is associated with a more benign illness course which, though not hugely different, may still constitute a large enough difference to make it relevant clinically, we must (1) make sure that the ethnic composition of the comparison groups matches the hypothesis, and (2) make sure we have a large enough sample to be able to detect it.