The concern is that the non respondents might have had more or less larynx cancer than the respondents. There is no particular reason to believe that they did, and hence no real reason to suspect the population surveyed is not representative of the entire cohort. One could speculate that more larynx cancers may have occurred among decedents, who were underrepresented in the survey. If this were true, the survey population might have somewhat lower larynx cancer rates than the total cohort. Such a bias is called a "conservative" bias, and would only make a positive finding in the study more impressive.