The Reply

We thank Singh and Grossman et al for their thoughtful comments. In response to Singh’s first point,we acknowledge that in ideal experimental settings,association proves causation. Causal inference from observational designs,on the other hand,can be challenging. While the advantages to performing studies using large databases are far-reaching,secondary analyses are often limited by the quality and completeness of data. For our study,an ideal analysis would have been to incorporate thiazide exposure as a time-dependent covariate,and would have included the 1055 patients who did not receive ongoing treatment. Unfortunately,the database we used was limited by incomplete data on the quantity of medication dispensed with each prescription. Consequently,we could not determine the exact length of medication exposure for each patient. In order to ensure that the patients analyzed had ongoing exposure,we limited our cohort to those who continued to receive antihypertensive prescriptions until the end of follow-up. Therefore,our results apply only to patients who receive treatment with thiazides and continue to take them until the development of hyponatremia or death.
The American Journal of Medicine

Be Sociable,Share!

•