Invited commentary

Abstract

In this study, a large and well-developed database and powerful statistical methods are used to construct a model describing a relationship between nadir hematocrit (Hct) on bypass and a decrement in postoperative renal function. Because of statistical power, this type of investigation has the potential to supersede the effort of individual investigators. That said, the intrinsic limitations of database studies are considerable and must be appreciated before results are embraced.It is not surprising that low perioperative Hct is associated with a decrement in renal function. However, the methods employed are not sufficient to determine if low Hct is itself the cause of decreasing renal function or if the low Hct during bypass is an expression of comorbid conditions such as perioperative blood loss, chronic anemia, CHF, MI, or acuity of operation. While a “co-morbidity index” was applied it did not include any of these factors.Equally important, the statistical model did not separate clinically meaningful from nonclinically, meaningful decrements in renal function. As such, a change in creatinine from 0.7 to 1.1 mg/dL was weighted equally to a change from 1.7 to 2.7 mg/dL. Similarly, the statistical modeling was unable to identify a Hct threshold below, which a meaningful decrement of renal function might be expected; the model showed only that higher Hcts are less likely to be associated with an increase in creatinine.Finally, as in most database studies, associations can be identified, but not mechanisms. Other literature strongly suggest that the primary determinant of post-cardiac surgical renal insufficiency is preexisting, often occult, renal disease, combined with perioperative low cardiac output. In this broader context it is difficult to understand the relative importance of a brief period (bypass) of reduced Hct. The statistical description of even a large database does not necessarily provide much clarification. In this study, a large and well-developed database and powerful statistical methods are used to construct a model describing a relationship between nadir hematocrit (Hct) on bypass and a decrement in postoperative renal function. Because of statistical power, this type of investigation has the potential to supersede the effort of individual investigators. That said, the intrinsic limitations of database studies are considerable and must be appreciated before results are embraced. It is not surprising that low perioperative Hct is associated with a decrement in renal function. However, the methods employed are not sufficient to determine if low Hct is itself the cause of decreasing renal function or if the low Hct during bypass is an expression of comorbid conditions such as perioperative blood loss, chronic anemia, CHF, MI, or acuity of operation. While a “co-morbidity index” was applied it did not include any of these factors. Equally important, the statistical model did not separate clinically meaningful from nonclinically, meaningful decrements in renal function. As such, a change in creatinine from 0.7 to 1.1 mg/dL was weighted equally to a change from 1.7 to 2.7 mg/dL. Similarly, the statistical modeling was unable to identify a Hct threshold below, which a meaningful decrement of renal function might be expected; the model showed only that higher Hcts are less likely to be associated with an increase in creatinine. Finally, as in most database studies, associations can be identified, but not mechanisms. Other literature strongly suggest that the primary determinant of post-cardiac surgical renal insufficiency is preexisting, often occult, renal disease, combined with perioperative low cardiac output. In this broader context it is difficult to understand the relative importance of a brief period (bypass) of reduced Hct. The statistical description of even a large database does not necessarily provide much clarification.