Advertisement

Methodologic Issues Relating to Outcomes in Severe Hyponatremia With and Without Desmopressin

      To the Editor:
      We read the article by MacMillan and Cavalcanti with great interest.
      • MacMillan T.E.
      • Cavalcanti R.B.
      Outcomes in severe hyponatremia treated with and without desmopressin (DDAVP).
      They proposed to compare outcomes in hyponatremia on the basis of desmopressin (DDAVP) usage. They found that baseline characteristics are statistically different between the groups with and without DDAVP; but we wonder why they compared outcomes between these 2 groups using univariable models and why no confounders were controlled between the groups. Additionally, it should be stated that the number of patients in each group was high, and many confounders could be controlled in their study. Hence, their results may be biased owing to confounding effects.
      • Rothman K.J.
      • Greenland S.
      • Lash T.L.
      Modern Epidemiology.
      Our second point is that the authors stated that the quantitative variables were compared by using Wilcoxon and Kruskal-Wallis tests, which is problematic. The Wilcoxon test is used when nonnormally distributed quantitative variables are compared in related or dependent samples, whereas the samples in their study were independent and the Mann-Whitney U test should be used to obtain valid results.
      • Cleophas T.J.
      • Zwinderman A.H.
      Non-Parametric Tests. Statistical Analysis of Clinical Data on a Pocket Calculator.
      Finally, the multivariable logistic regression model was used to predict odds of in-hospital death, but confounders were not selected according to the defined criteria. It has been suggested that variables for which the estimated effect size (odds ratio in their study) is changed more than 10% in the univariable and corresponding multivariable model should be selected as confounders and imported into the multivariable model.
      • Lee P.H.
      Is a cutoff of 10% appropriate for the change-in-estimate criterion of confounder identification?.
      Otherwise nonconfounders may be imported into multivariable model and lead to over-parameterization of the model; and on the other hand, some confounders may be missed.
      The take-home message for readers is that multivariable regression models should be applied when the distribution of confounders is different between studied groups. Additionally, the confounders should be detected according to the standard criteria to obtain a parsimonious model with efficient controlling of confounders. Finally, nonparametric tests should be used appropriately considering the dependency of samples.

      References

        • MacMillan T.E.
        • Cavalcanti R.B.
        Outcomes in severe hyponatremia treated with and without desmopressin (DDAVP).
        Am J Med. 2017; https://doi.org/10.1016/j.amjmed.2017.09.048
        • Rothman K.J.
        • Greenland S.
        • Lash T.L.
        Modern Epidemiology.
        Lippincott Williams & Wilkins, Philadelphia2008
        • Cleophas T.J.
        • Zwinderman A.H.
        Non-Parametric Tests. Statistical Analysis of Clinical Data on a Pocket Calculator.
        Springer, New York2011: 9-13
        • Lee P.H.
        Is a cutoff of 10% appropriate for the change-in-estimate criterion of confounder identification?.
        J Epidemiol. 2014; 24: 161-167

      Linked Article

      • The Reply
        The American Journal of MedicineVol. 131Issue 6
        • Preview
          We thank Drs. Safiri and Ashrafi-Asgarabad for their interest in our study on outcomes in severe hyponatremia with desmopressin (DDAVP) usage.1 They raise 3 issues about the methodology.
        • Full-Text
        • PDF