It's not restricted. NIH won't allow you to use the datasets they control for it. It is practically and pragmatically reasonable for them to do that. It's also ethically sound to not let the data subjects data be used for something they wouldn't want as the data controller.
If a scientist wants to do the research then they can pursue it and find a journal to publish it in. The reputable journals are unlikely to find it of interest unless it was a huge effect size. But we know any effect size must be tiny hence the need for a huge dataset.
They'd have to take samples and basically tell the person the intension was to check if some races were genetically inferior let's see how well they do with recruitment. I wouldn't give a sample.
Or just do a GWAS study and then develop metrics for estimating risk for individuals which can make use of ethnicity.
If a scientist wants to do the research then they can pursue it and find a journal to publish it in. The reputable journals are unlikely to find it of interest unless it was a huge effect size. But we know any effect size must be tiny hence the need for a huge dataset.
They'd have to take samples and basically tell the person the intension was to check if some races were genetically inferior let's see how well they do with recruitment. I wouldn't give a sample.
Or just do a GWAS study and then develop metrics for estimating risk for individuals which can make use of ethnicity.