Extrapolation of reciprocal creatinine plot is not reliable in predicting the onset of dialysis in patients with progressive renal insufficiency

Background. Reciprocal creatinine plot is often used to monitor patients with progressive renal insufficiency and to predict the onset of dialysis, although the latter practice has not been validated. Objective. We examined whether extrapolating the reciprocal creatinine plot can predict the onset of dialysis. Setting. Single centre study in the dialysis unit of a University teaching hospital. Design. We studied 170 consecutive patients with progressive renal insufficiency referred to a single nephrology unit and subsequently dialysed. Reciprocal creatinine plot was constructed by all available serum creatinine values before dialysis (the 'definitive plot'). Four 'interim plots' were constructed for each patient by using serum creatinine below 400, 500, 600 and 700 mumol L-1 . Interim plots with at least five points and Pearson's r > 0.9 were analysed. The date of dialysis was predicted from the least squares linear regression formula and a target serum creatinine level cor- responding to estimated creatinine clearance of 7 mL min(-1) , at which dialysis was recommended. Results. The median duration of observation was 25 months. After serum creatinine 500 mumol L-1 , the slope of the interim plot remained stable and extrapolation was possible in 117 patients (68.8%). However, the limits of agreement for predicting the onset of dialysis were wide (from -11.7 to +9.5 months). At this creatinine level, the onset of dialysis fell within 1 month of the predicted onset in only 41 patients (24.1%). The limits of agreement for prediction narrowed when time points of higher serum creatinine were included into the plot. However, nine patients (5.3%) required dialysis within 1 month at creatinine 600 mumol L-1 and the dialysis was not predicted by the reciprocal creatinine plot. Target serum creatinine did not correlate with acute serum creatinine at which dialysis was started (r = 0.051, P = 0.51). A slower decline in renal function was associated with a higher prediction error (r = 0.212, P = 0.014). Conclusions. The onset of dialysis cannot be predicted by extrapolation of the reciprocal creatinine plot because of individual variation in the renal function that require dialysis. Dialysis would be almost imminent in some patients by the time serum creatinine reaches a level that allows accurate construction and extrapolation of a plot.