Emergency medicine pharmacotherapy compromises accuracy of plasma creatinine determination by enzyme-based methods: real-world clinical evidence and implications for clinical practice

Abstract

Background: Assessment of kidney function in emergency settings is essential across all medical subspecialties. Daily assessment of patient creatinine results from emergency medical services showed that some deviated from expected values, implying drug-related interference.

Methods: Real-time clinical evaluation of an enzyme method (Roche CREP2) in comparison with the Jaffé gen. 2 method (Roche CREJ2) was performed. During the period of December 2022 and January 2023, we analyzed 8,498 patient samples, where 5,524 were heavily medicated STAT patient specimens, 500 were pediatric specimens, and 2,474 were from a distant general population in a different region using the same methods.

Results: In 109 out of 5,524 hospital specimens (1.97%, p < 0.001), the CREP2 value was apparently (25% or more) lower than CREJ2. Suspect interfering medication was found in a sample of 43 out of 46 reviewed patients where medication data were available. This phenomenon was not observed in the general population.

Conclusion: In a polymedicated urgent care hospital population, a creatinine enzyme method produces unreliable results, apparently due to multiple drug-related interferences.

Keywords: creatinine; cystatin C; eGFR; renal functions; urgent care medication.

Figure 1

CREP2 exhibits a strong susceptibility to underestimate creatinine concentrations in the “hospital adult urgent care population” in contrast to the “common adult population.” (A) Logarithmic-scale scatterplot with paired data for the dataset “hospital adult urgent care population” with the samples being rank-ordered by creatinine concentrations determined by the CREJ2 method. The data show that CREP2 values were much lower than CREJ2 values in approximately 2% of the samples. (B) Logarithmic-scale scatterplot with identity line for the same dataset as in panel (A). The identity line represents perfect agreement between the two methods; no negative interferences in the urgent care cohort with the CREJ2 method were observed. (C) Logarithmic-scale scatterplot with identity line for the dataset “common adult population” showing agreement between the two methods. (D) Bland–Altman plot for dataset “hospital adult urgent care population” from panel (A); the horizontal solid line represents the mean difference between the two values, whereas the red (dotted) lines represent the limits of agreement (1.96 SD) of individual differences. (E) Bland–Altman plot for the same dataset as in panel (D) where outlying observations greater than 25% were removed. (F) Bland–Altman plot for dataset “common adult population” where no deviations between the two methods in the general community patient cohort were observed.

Figure 2

Pediatric hospital population does not exhibit susceptibility to underestimate creatinine concentrations by CREP2. (A) Bland–Altman plot for dataset “pediatric hospital population” showing 25 of 500 values greater than 1.96 SD. These cases were then individually screened by a pediatric nephrologist (JS) and a specialist in pediatric laboratory medicine (DV) for suspect medication. (B) Bland–Altman plot for the same dataset with any samples exhibiting the difference between the two methods greater than 25% removed.

Figure 3

Time-dependent individual patient trends of consecutive creatinine measurements expressed as CREP2/CREJ2 ratio and aggregated and ranked to “sampling time events” from the first event through the tenth event. To visualize individual time-dependent trends of ratios of creatinine concentrations, we used relative timescale aggregating measurements as ordinal data. Boxplots of all patients/all measurements showed as “sampling time events 1 through 10” on the x-axis against the ratio of creatinine concentrations by CREP2/CREJ2 are shown. Major deviations from the median line observed among the 2nd through 7th measurement events implying a drug effect present but then returning to result concordance showed as a ratio approximately 1 implying a drug effect being “washed out.”