Pharmacometric modelling of risk factors for MRSA

We identified all patients (pts) with C&S showing susceptible Staphylococcus aureus (SA), seen from 10/92 →10/95, using a computerized database. Of 1160 pts, 38 subsequently had methicillin resistant SA (MRSA) isolated, 1122 remained susceptible. Pharmacometric models were developed (maximum likelihood, ADAPT II; model discrimination by Akaike's Information Criterion) to identify factors that were significantly associated with risk of subsequent MRSA. In an initial univariate screen, factors predictive of MRSA included low AUIC (antimicrobial AUC₂₄/MIC), ICU as treatment site, fluoroquinolone (FQ) or cefazolin usage & blood, respiratory or urine as site of infection. In the final pharmacometric model, two dichotomous (ICU & FQ) & one continuous factor (AUIC), remained significant. A separate Hill-type function, defining the relationship between AUIC & percent probability (%P) for MRSA, was fit to each of the 4 pt groups (ICU &/or FQ, Yes/No): %P = Po[1-Emax-AUIC^H/(AUICm^H+AUIC^H)], in which Po is the extrapolated %P at AUIC=0, Emax is the maximum AUIC-related reduction in %P, AUICm is the AUIC at which the effect is 1/2Emax & H is Hill's constant. Fitted values for H were > 10 in all 4 groups, other fitted values included: Group 1 Group 2 Group 3 Group 4 Po% 20.5 13.5 28.1 87.3 Emax% 91.1 83.1 90.8 82.9 AUICm 40.0 113 60.5 95.7 Groups 1 (ICU/FQ, no/no; n = 662), 2 (ICU/FQ, yes/no; n = 187) & 3 (ICU/FQ, no/yes; n = 273) had similar %Po & %Emax, but differed in the AUIC needed to reduce %P; at AUIC > AUICm the %P was 2-3%. Group 4 (ICU/FQ, yes/yes; n = 38) had higher %P overall; at AUIC < 90, %P≈85%; at AUIC > 150, %P≈15%. There was a highly significant relationship between AUIC & %P; optimization of drug regimens for AUIC should substantially reduce incidence of MRSA.