Although the breadth of variables included in NHANES enabled us to examine and adjust for many potential confounders, residual confounding and possible overadjustment could be sources of bias.
We adjusted for many potential confounders, but residual confounding cannot be discarded.
Moreover, although we adjusted for many potential confounders, residual and unmeasured confounding is a possibility.
However, despite adjustment for many potential confounders in multivariable models, residual confounding (e.g. by other macronutrients, fiber or flavonoid intake) in observational studies cannot fully be excluded.
Even though we have adjusted for many potential confounders, we cannot exclude residual confounding or reverse causation.
Finally, the observed associations could be due to unmeasured confounding, although we investigated many potential confounders.
However, also for these studies, despite inclusion of many potential confounders in their multivariate model, residual confounding (e.g. by intake of other macronutrients or salt) cannot be excluded.
Although we collected information on many potential confounders, there was little evidence of actual confounding by the factors evaluated.
First, although many potential confounders had been adjusted in the models, residual confounding cannot be excluded.
Fifthly, despite adjustment for many potential confounders through a propensity score technique, residual and unmeasured confounding remains possible as our study is observational.
33 Finally, although the associations have been adjusted for many potential confounders including smoking, drug use and physical inactivity, residual confounding cannot be ruled out in observational studies.
