Table 1.

Input parameters in simulations

CategoryFactorValues Considered in Simulations
GFR declineMean long-term slope−1.5, -3.25, or −5 ml/min per 1.73 m2
SD of long-term slope3.0, 3.5, 4.0, or 4.5 ml/min per year/1.73 m2
Correlation of slope and intercept−0.03a
Slope skewnessSlight negative skewness (generalized log γ shape parameter=3)a
Autocorrelation0
Size of residualsResidual variance=0.67×expected GFR (low variability) or 0.817×expected GFR (moderate variability)a
Acute effectMean acute effect−2.5, −1.25, 0, +1.25, or 2.5 ml/min per 1.73 m2 at baseline GFR=42.5 ml/min per 1.73 m2
Attenuation of initial acute effectLinear to 15 ml/min per 1.73 m2 or no attenuation
Variability of acute effectAcute effect SD=1 ml/min per 1.73 m2
Long-term treatment effectType of long-term treatment effectProportional, uniform or intermediateb
Size of long-term treatment effect0% or 25%, reduction in the slope for a subject with an average long-term slope in the absence of treatment
Death and ESKDDeath rate per year0.03375–0.00025×expected GFR (ranges from 0.030 at GFR=15 to 0.01125 at GFR=90 ml/min per 1.73 m2)a
GFR level associated with onset of ESKDUniformly distributed between 6 and 15 ml/min per 1.73 m2
Design and conductAccrual period and total follow-upShort: 1 yr accrual and 1.5 yr further follow-up
Medium: 1.5 yr accrual and 2.5 yr further follow-up
Long: 2 yr accrual and 4 yr further follow-up
Measurement frequencyMonths 3, 6, and every 6 mo thereafter
Mean baseline GFR27.5, 42.5, or 67.5 ml/min per 1.73 m2
No. of baseline GFRs2
Permanent loss to follow-up rate2%/yr
Intermittent missing GFRs5%
  • a See supplement to reference 8 for additional details.

  • b Under the proportional effect model, the distribution of chronic GFR slopes in the treatment group were generated by simulating the slope of the ith patient as βi×{(1−k)1[βi<0]+1[βi≥0]}, where βi denotes the slope the patient would have had without the treatment, k is the proportional reduction due to the treatment among patients whose GFR would have declined without the treatment, and 1[βi<0] and 1[βi≥0] are 0–1 indicator variables for negative and non-negative slopes, respectively. Thus the proportional effect model assumes the treatment reduces the magnitude of the slope by 100×k percent among patients whose slope would have been negative without the treatment but has no effect on patients whose slope would have been ≥0 without the treatment. Under the uniform treatment effect model the chronic GFR slopes in the treatment group were generated as βi−k×mean(βi) where mean(βi) is the mean chronic slope without the treatment. Under the intermediate treatment effect model, the chronic slopes in the treatment group were generated as [βi−(k/2)×mean(βi)]+[βi×{[1−(k/2)]1[βi<0]+1[βi≥0]}].