Author's response

We would like to thank authors for their comments on our article,1 which are both valid and valuable to consider. We agree that the estimated odds ratios are biased away from the null hypothesis given the sparse data. We also agree that the use of the word independently is incorrect in the abstract. The odds ratios reported in the abstract and in table 4 are univariate odds ratios. The term independently associated should be reserved for results from a multivariable analysis in which we have adjusted for the other variables in the model.

http://ift.tt/2ePUSjl