OBJECTIVE: This study was performed to determine prospectively whether, in the presence of proved or presumed bacterial infection, the sensitivity of serum C-reactive protein (CRP) response could be enhanced by serial rather than single determinations. We also sought to assess CRP responses to clinically identified noninfectious disorders. DESIGN: The CRP responses of 491 infants on 691 occasions of suspected infection were assessed. CRP levels were measured initially and twice again at 12-hour intervals (rate immunonephelometry). Assessments also included a blood culture, complete blood cell count, and chest radiograph and culture of spinal fluid when appropriate. CRP responses were correlated with four designated clinical groups: (1) positive blood or cerebrospinal fluid cultures (n = 190); (2) negative blood culture-definite infection (necrotizing enterocolitis stages 2 and 3, pneumonia, subcutaneous abscess) (n = 52); (3) negative blood culture-possible infection (antenatal risk factors, meconium aspiration, positive urine group B streptococcus antigen, necrotizing enterocolitis stage 1, febrile infants) (n = 287); and (4) negative blood culture-no infection (respiratory distress syndrome, transient tachypnea of the newborn, patent ductus arteriosus, tissue trauma) (n = 160). Diagnoses were made before CRP results were known. RESULTS: In all, 187 (27%) of the blood cultures were positive. A single organism was recovered from 174 of these; two organisms from 13. Among the single-organism cultures, 50 (29%) were Gram-negative, 120 (69%) were Gram-positive, and 4 (2%) were budding yeasts. CRP levels were elevated in various groups as follows: in the positive blood culture group (by organism), Gram-negative rods, 92% (46/50); group B streptococcus, 92% (12/13); Staphylococcus aureus, 89% (8/9); group D streptococcus, 71% (10/14); Streptococcus viridans, 60% (6/10); Staphylococcus epidermidis, 55% (40/73). In the negative blood culture-definite infection group, CRP levels were abnormal in 88%; in the negative culture-possible infection group, CRP was elevated in 33%; and in the negative blood culture-no infection group, CRP was elevated in 9%. Serial determinations of CRP resulted in enhanced sensitivity in the positive blood culture group, the negative blood culture-definite infection group, and the negative blood culture-possible infection group. Initial determinations by themselves were inadequately sensitive. Serial determinations did not enhance sensitivity of the negative blood culture-no infection group. High specificity (91%) is suggested by the low incidence of abnormal CRP levels among infants who were not infected. CONCLUSIONS: These data suggest that it would be appropriate to conduct a cautious, controlled trial to assess the safety of discontinuing antibiotic therapy if three serial CRP measurements are normal and if there are no other clinical factors suggestive of infection. The data also indicate the necessity for serial determinations of CRP for optimal sensitivity.
OBJECTIVE: This study was performed to determine prospectively whether, in the presence of proved or presumed bacterial infection, the sensitivity of serum C-reactive protein (CRP) response could be enhanced by serial rather than single determinations. We also sought to assess CRP responses to clinically identified noninfectious disorders. DESIGN: The CRP responses of 491 infants on 691 occasions of suspected infection were assessed. CRP levels were measured initially and twice again at 12-hour intervals (rate immunonephelometry). Assessments also included a blood culture, complete blood cell count, and chest radiograph and culture of spinal fluid when appropriate. CRP responses were correlated with four designated clinical groups: (1) positive blood or cerebrospinal fluid cultures (n = 190); (2) negative blood culture-definite infection (necrotizing enterocolitis stages 2 and 3, pneumonia, subcutaneous abscess) (n = 52); (3) negative blood culture-possible infection (antenatal risk factors, meconium aspiration, positive urine group B streptococcus antigen, necrotizing enterocolitis stage 1, febrile infants) (n = 287); and (4) negative blood culture-no infection (respiratory distress syndrome, transient tachypnea of the newborn, patent ductus arteriosus, tissue trauma) (n = 160). Diagnoses were made before CRP results were known. RESULTS: In all, 187 (27%) of the blood cultures were positive. A single organism was recovered from 174 of these; two organisms from 13. Among the single-organism cultures, 50 (29%) were Gram-negative, 120 (69%) were Gram-positive, and 4 (2%) were budding yeasts. CRP levels were elevated in various groups as follows: in the positive blood culture group (by organism), Gram-negative rods, 92% (46/50); group B streptococcus, 92% (12/13); Staphylococcus aureus, 89% (8/9); group D streptococcus, 71% (10/14); Streptococcus viridans, 60% (6/10); Staphylococcus epidermidis, 55% (40/73). In the negative blood culture-definite infection group, CRP levels were abnormal in 88%; in the negative culture-possible infection group, CRP was elevated in 33%; and in the negative blood culture-no infection group, CRP was elevated in 9%. Serial determinations of CRP resulted in enhanced sensitivity in the positive blood culture group, the negative blood culture-definite infection group, and the negative blood culture-possible infection group. Initial determinations by themselves were inadequately sensitive. Serial determinations did not enhance sensitivity of the negative blood culture-no infection group. High specificity (91%) is suggested by the low incidence of abnormal CRP levels among infants who were not infected. CONCLUSIONS: These data suggest that it would be appropriate to conduct a cautious, controlled trial to assess the safety of discontinuing antibiotic therapy if three serial CRP measurements are normal and if there are no other clinical factors suggestive of infection. The data also indicate the necessity for serial determinations of CRP for optimal sensitivity.