Spontaneous changes in the rate of decline in reciprocal serum creatinine: errors in predicting the progression of renal disease from extrapolation of the slope.

The slope of reciprocal serum creatinine (1/Pcr) versus time has been used to measure the rate of progression of chronic renal disease, predict the interval until onset of end-stage renal disease, and assess the effect of therapy. In order to determine the errors that might result from extrapolating the slope of 1/Pcr versus time beyond the interval of observation, we applied a method of linear regression analysis to search for spontaneous changes in slope in 21 patients from New England Medical Center and 56 patients in three published studies in whom the decline in 1/Pcr appeared constant (r greater than or equal to 0.84 for the correlation of 1/Pcr versus time). Significant changes in the slope (breakpoints) were identified in one third to one half of the 77 patients and appeared to be spontaneous. The second slope was less steep in 49 patients (6.1%); the mean value for serum creatinine at the time of the breakpoint was 5.3 mg/dL; the mean change in slope (absolute value) was 0.005 dL/mg/month (adults) and 0.017 dL/mg/month (children); and the mean error in prediction of the interval until the final value for serum creatinine was 27% of the actual interval. We conclude that spontaneous breakpoints in the slope of 1/Pcr versus time are very frequent, even among patients with an apparent constant rate of decline. Breakpoints may cause errors in extrapolating the slope to predict the interval until the onset of end-stage renal disease and to assess the effect of therapy.(ABSTRACT TRUNCATED AT 250 WORDS)