Stepwise multivariable logistic regression analysis for bacterial predictors of fecal or pyelonephritis source among 329 Escherichia coli isolates from women with cystitis or pyelonephritis and uninfected womena

Outcome variable (referent)Predictor variablebP valueOdds ratio95% CI
Fecal (cystitis or pyelo) iha 0.0040.340.17-0.71
papEF <0.0010.270.14-0.54
ibeA <0.0010.160.06-0.44
Pyelo (fecal or cystitis) IbGroup B20.0011.371.14-1.64
papG allele II<0.0016.303.71-10.70
Pyelo (fecal or cystitis) IIb papG allele II<0.0017.654.43-13.21
  • a Pyelo, pyelonephritis; CI, confidence interval; afa/draBC, Dr-binding adhesins; ibeA, invasion of brain endothelium; iha, putative adhesin siderophore; K1, K1 capsule variant; papEF, P fimbriae minor pilins; papG allele II, P fimbriae adhesin variant; sfa/focDE, S and F1C fimbriae.

  • b Predictor variables shown are those from the last step in which all variables included in the model yielded a P value of ≤0.015. Two different sets of candidate predictor variables were used for each endpoint (I, full set; II, short set). The first (full) set included all bacterial characteristics analyzed, i.e., the four phylogenetic groups, the UTI-associated O antigens (individually and collectively), the O11/O17/O73/O77 antigens, the individual papA alleles, individual virulence factors, and ExPEC status. The second (short) set included only the papA alleles, individual virulence factors, and ExPEC status. Results for “Fecal (cystitis or pyelo)” were the same with either set of candidate predictor variables. For “Pyelo (fecal or cystitis),” the two sets of candidate predictor variables yielded different results, as shown.