Preoperative Dialysis Dose and Postoperative Outcomes in Patients Receiving Maintenance Hemodialysis

Abstract

Key Points:

1. Little is known about preoperative hemodialysis dosing for patients with ESKD.

2. Among patients receiving maintenance hemodialysis, preoperative decreases in Kt/V_urea were significantly associated with postoperative mortality.

3. The connection between preoperative hemodialysis practices and postoperative outcomes warrants further investigation.

Background: Little is known about preoperative hemodialysis dosing for patients with ESKD. We assessed whether changes in preoperative hemodialysis dose (compared with and controlling for baseline dose) are associated with postoperative mortality in patients receiving maintenance hemodialysis.

Methods: We identified fee-for-service Medicare beneficiaries receiving hemodialysis for ESKD who underwent surgical procedures between January 1, 2011, and November 30, 2020. Follow-up ended December 31, 2020. The primary exposure was preoperative change in Kt/V_urea, defined as the difference between the Kt/V_urea in the hemodialysis session proximal to the procedure and the mean Kt/V_urea for the preceding 180 days. The primary outcome was postoperative 30-day mortality. The relation between preoperative change in Kt/V_urea and the primary outcome was modeled using a Cox proportional hazards regression model, adjusted for mean and SD of Kt/V_urea in the 180 days preceding the procedure, and for other covariates.

Results: Among 151,240 procedures (median age, 65 years [25%–75% range, 56–73], 63,437 [41.9%] in women), 31,825 (21.0%) had a preoperative change in Kt/V_urea of <−0.10, 43,790 (29.0%) had a preoperative change of −0.10 to <0, 45,058 (29.8%) had a preoperative change of 0 to <+0.10, and 30,567 (20.2%) had a preoperative change of ≥+0.10. The median Kt/V_urea for the 180 days before the procedure was 1.58 (25th–75th percentiles, 1.45–1.74). In adjusted analysis, compared with patients with a preoperative change in Kt/V_urea of 0 to <+0.10, 30-day mortality was 1.50 (95% confidence interval, 1.32 to 1.70) times higher with a preoperative Kt/V_urea change of ≤−0.10 and 1.16 (95% confidence interval, 1.02 to 1.31) times higher with a preoperative Kt/V_urea change of −0.10 to <0. Increases in preoperative Kt/V_urea that were >0.10 were not significantly associated with 30-day mortality.

Conclusions: Among Medicare beneficiaries receiving maintenance hemodialysis, preoperative decreases in Kt/V_urea (compared with and controlling for mean Kt/V_urea) were significantly associated with postoperative mortality.

Keywords: USRDS (United States Renal Data System); hemodialysis; hemodialysis adequacy; outcomes.

Graphical abstract

Figure 1

Unadjusted 30-day postoperative mortality, by change in preoperative Kt/V from 180-day baseline Kt/V (ΔKt/V). This figure shows the unadjusted cumulative incidence of mortality by preoperative change in Kt/V (ΔKt/V) for the entire study cohort. ΔKt/V is defined as the difference between the 180-day mean Kt/V and the Kt/V measured in the hemodialysis session immediately prior (within 3 days) of the surgical procedure. For example, ΔKt/V <−0.10 signifies that the preoperative Kt/V was more than 0.1 lower than the 180-day mean for the patient for a given procedure. Log-rank P < 0.001 for ΔKt/V <−0.10 compared with ΔKt/V 0 to <0.10. Log-rank P = 0.01 for ΔKt/V −0.10 to <0 compared with ΔKt/V 0 to <0.10. Log-rank P = 0.001 for ΔKt/V ≥+0.10 compared with ΔKt/V 0 to <0.10.

Figure 2

Association between change in preoperative Kt/V from 180-day baseline Kt/V (ΔKt/V) and postoperative outcomes. This figure reports the adjusted association between ΔKt/V and postoperative outcomes. ΔKt/V is defined as the difference between the 180-day mean Kt/V and the Kt/V measured in the hemodialysis session immediately prior (within 3 days) of the surgical procedure. For example, ΔKt/V <−0.10 signifies that the immediately preoperative Kt/V was more than 0.1 lower than the 180-day mean for the patient for a given procedure. All models adjusted for covariables described in the text and in Table 1, including the 180-day mean Kt/V for the patient as well as the variability of Kt/V in the 180 days preceding the surgery in an effort to account for dose-targeting bias. aHR, adjusted hazard ratio; ARD, absolute risk difference; CI, confidence interval.

Figure 3

Association between change in preoperative Kt/V from 180-day baseline Kt/V (ΔKt/V) and 30-day postoperative mortality. This figure displays results of an analysis of the adjusted association between ΔKt/V and 30-day postoperative mortality, with ΔKt/V modeled using a restrictive cubic spline with knots at 5%, 27.5%, 50%, 72.5%, and 95% (dashed vertical lines). Estimated HRs (solid line) and 95% CIs (shaded area) for 30-day postoperative mortality are displayed across the range of ΔKt/V values on the x axis, with a ΔKt/V of 0 (not change in preoperative Kt/V from 180-day mean Kt/V) used as the reference level. Similar to previous figures, ΔKt/V is defined as the difference between the 180-day mean Kt/V and the Kt/V measured in the hemodialysis session immediately prior (within 3 days) of the surgical procedure. For example, ΔKt/V <−0.10 signifies that the immediately preoperative Kt/V was more than 0.1 lower than the 180-day mean for the patient for a given procedure. All models adjusted for covariables described in the text and in Table 1, including the 180-day mean Kt/V for the patient as well as the variability of Kt/V in the 180 days preceding the surgery in an effort to account for dose-targeting bias. HR, hazard ratio.

Figure 4

Association between change in preoperative Kt/V from 180-day baseline Kt/V (ΔKt/V) and 30-day postoperative mortality, by 180-day baseline Kt/V. This figure displays results of an analysis exploring whether the association between preoperative ΔKt/V and postoperative mortality depends on the patient's 180-day mean Kt/V. In each panel, 180-day mean Kt/V from 1.1 to 2.2 is displayed on the x axis, representing the first percentile to 99th percentile range of all mean Kt/V values in the study. Estimated HRs or ARDs (solid line) and 95% CIs (shaded area) for large decrease in Kt/V (ΔKt/V <−0.10) (A), small decreases in Kt/V (ΔKt/V −0.10 to <0) (B), and large increases in Kt/V (ΔKt/V ≥+0.10) (C) compared with the reference category of no change or a small increase in Kt/V (ΔKt/V 0 to <0.10) are displayed across the range of 180-day mean Kt/V values. Similar to previous figures, ΔKt/V is defined as the difference between the 180-day mean Kt/V and the Kt/V measured in the hemodialysis session immediately prior (within 3 days) of the surgical procedure. For example, a large decrease in Kt/V (ΔKt/V <−0.10) signifies that the preoperative Kt/V was more than 0.1 lower than the 180-day mean for the patient for a given procedure. Mean Kt/V was modeled using a restrictive cubic spline with four knots at the 5th, 35th, 65th, and 95th percentiles. Multiplicative interaction terms were used to model the interaction between preoperative ΔKt/V and the restricted cubic spline representing 180-day mean Kt/V. In addition to these terms, all models were adjusted for covariables described in the text and in Table 1, including the variability of Kt/V in the 180 days preceding the surgery.