Table 7.

Sensitivity analyses by serum digoxin level

ParameterNUnadjusted Model (HR [95% CI])Covariate- and Propensity Score–Adjusted Model (HR [95% CI])a
Mortality risk by time-varying serum digoxin level
    per 1-ng/ml increase3313b1.22 (1.11 to 1.34)1.11 (1.00 to 1.24)
    no digoxin level monitoring40.7%c1.16 (1.02 to 1.32)1.14 (0.99 to 1.32)
    <0.6 ng/ml23.8%1.00 (reference)1.00 (reference)
    0.6 to 0.9 ng/ml12.6%1.05 (0.89 to 1.24)1.06 (0.87 to 1.29)
    1.0 to 2.2 ng/ml20.3%1.33 (1.16 to 1.53)1.19 (0.98 to 1.44)
    >2.2 ng/ml2.6%1.59 (1.14 to 2.23)1.14 (1.01 to 1.30)
    P for trend0.00070.14
Mortality risk after covariate matching
    matched on facility9281.22 (1.03 to 1.43)1.05 (0.88 to 1.27)
    matched on physician8401.27 (1.08 to 1.50)1.18 (1.01 to 1.38)
    matched on propensity scored16261.22 (1.09 to 1.37)1.55 (1.05 to 2.27)
  • aParameters used in the propensity score were the same as the primary analysis (see Table 2 footnote); covariates used in the Cox model were the same as the primary analysis with the exception of serum digoxin level, K, calcium, dialysate potassium, dialysate calcium, albumin, hemoglobin, systolic and diastolic BP (pre- and postdialysis readings), weight, interdialytic weight gain, and Charleson comorbidity index, which were modeled as time-dependent parameters (90-day cycles).

  • bA total of 1237 of 4550 patients were censored at death because no in-center serum digoxin level monitoring was done.

  • cResults are reported as a percentage of time in each serum digoxin ordinal given patients had changing serum digoxin levels during the follow-up period.

  • dStandardized difference = 0.4% and variance ratio = 0.02% for the propensity score between the groups after matching.