Personal exposure assessment studies may suffer from exposure-relevant selection bias