Continued controversy in using latent class models for estimating diagnostic accuracy without a gold standard.